A
1
nyNPU-RTL/Master/gemma3NE2B_int4_quantization.py
2
총 3개의 safetensors 파일을 변환 시작합니다...
3
4
[model-00002-of-00003.safetensors] 처리 중...
5
-> [양자화 O] model.audio_tower.conformer.0.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
6
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
7
-> [양자화 O] model.audio_tower.conformer.0.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
8
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
9
-> [양자화 O] model.audio_tower.conformer.0.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
10
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
11
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.attention.post_norm.weight : 원본 형태 torch.Size([1536])
12
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
13
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
14
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
15
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
16
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
17
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
18
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
19
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
20
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
21
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
22
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
23
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
24
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
25
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
26
-> [양자화 X (원본유지)] model.audio_tower.conformer.0.norm.weight : 원본 형태 torch.Size([1536])
27
-> [양자화 O] model.audio_tower.conformer.1.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
28
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
29
-> [양자화 O] model.audio_tower.conformer.1.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
30
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
31
-> [양자화 O] model.audio_tower.conformer.1.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
32
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
33
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.attention.post_norm.weight : 원본 형태 torch.Size([1536])
34
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
35
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
36
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
37
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
38
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
39
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
40
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
41
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
42
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
43
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
44
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
45
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
46
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
47
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
48
-> [양자화 X (원본유지)] model.audio_tower.conformer.1.norm.weight : 원본 형태 torch.Size([1536])
49
-> [양자화 O] model.audio_tower.conformer.2.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
50
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
51
-> [양자화 O] model.audio_tower.conformer.2.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
52
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
53
-> [양자화 O] model.audio_tower.conformer.2.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
54
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
55
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.attention.post_norm.weight : 원본 형태 torch.Size([1536])
56
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
57
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
58
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
59
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
60
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
61
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
62
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
63
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
64
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
65
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
66
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
67
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
68
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
69
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
70
-> [양자화 X (원본유지)] model.audio_tower.conformer.2.norm.weight : 원본 형태 torch.Size([1536])
71
-> [양자화 O] model.audio_tower.conformer.3.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
72
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
73
-> [양자화 O] model.audio_tower.conformer.3.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
74
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
75
-> [양자화 O] model.audio_tower.conformer.3.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
76
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
77
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.attention.post_norm.weight : 원본 형태 torch.Size([1536])
78
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
79
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
80
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
81
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
82
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
83
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
84
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
85
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
86
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
87
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
88
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
89
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
90
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
91
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
92
-> [양자화 X (원본유지)] model.audio_tower.conformer.3.norm.weight : 원본 형태 torch.Size([1536])
93
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
94
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
95
-> [양자화 X (원본유지)] model.audio_tower.subsample_conv_projection.conv_0.conv.weight : 원본 형태 torch.Size([128, 1, 3, 3])
96
-> [양자화 X (원본유지)] model.audio_tower.subsample_conv_projection.conv_0.norm.weight : 원본 형태 torch.Size([128])
97
-> [양자화 X (원본유지)] model.audio_tower.subsample_conv_projection.conv_1.conv.weight : 원본 형태 torch.Size([32, 128, 3, 3])
98
-> [양자화 X (원본유지)] model.audio_tower.subsample_conv_projection.conv_1.norm.weight : 원본 형태 torch.Size([32])
99
-> [양자화 O] model.audio_tower.subsample_conv_projection.input_proj_linear.weight : 원본 형태 (1536, 1024)
100
-> [양자화 O] model.language_model.altup_projections.0.weight : 원본 형태 (2048, 2048)
101
-> [양자화 O] model.language_model.altup_projections.1.weight : 원본 형태 (2048, 2048)
102
-> [양자화 O] model.language_model.altup_projections.2.weight : 원본 형태 (2048, 2048)
103
-> [양자화 X (원본유지)] model.language_model.altup_unembed_projections.0.weight : 원본 형태 torch.Size([2048, 2048])
104
-> [양자화 X (원본유지)] model.language_model.altup_unembed_projections.1.weight : 원본 형태 torch.Size([2048, 2048])
105
-> [양자화 X (원본유지)] model.language_model.altup_unembed_projections.2.weight : 원본 형태 torch.Size([2048, 2048])
106
-> [양자화 X (원본유지)] model.language_model.embed_tokens_per_layer.weight : 원본 형태 torch.Size([262144, 7680])
107
-> [양자화 X (원본유지)] model.language_model.layers.26.altup.correct_output_scale : 원본 형태 torch.Size([2048])
108
-> [양자화 X (원본유지)] model.language_model.layers.26.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
109
-> [양자화 X (원본유지)] model.language_model.layers.26.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
110
-> [양자화 X (원본유지)] model.language_model.layers.26.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
111
-> [양자화 X (원본유지)] model.language_model.layers.26.altup.router_norm.weight : 원본 형태 torch.Size([2048])
112
-> [양자화 X (원본유지)] model.language_model.layers.26.input_layernorm.weight : 원본 형태 torch.Size([2048])
113
-> [양자화 X (원본유지)] model.language_model.layers.26.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
114
-> [양자화 X (원본유지)] model.language_model.layers.26.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
115
-> [양자화 X (원본유지)] model.language_model.layers.26.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
116
-> [양자화 O] model.language_model.layers.26.mlp.down_proj.weight : 원본 형태 (2048, 8192)
117
-> [양자화 X (원본유지)] model.language_model.layers.26.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
118
-> [양자화 X (원본유지)] model.language_model.layers.26.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
119
-> [양자화 X (원본유지)] model.language_model.layers.26.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
120
-> [양자화 X (원본유지)] model.language_model.layers.26.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
121
-> [양자화 X (원본유지)] model.language_model.layers.26.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
122
-> [양자화 X (원본유지)] model.language_model.layers.26.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
123
-> [양자화 X (원본유지)] model.language_model.layers.27.altup.correct_output_scale : 원본 형태 torch.Size([2048])
124
-> [양자화 X (원본유지)] model.language_model.layers.27.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
125
-> [양자화 X (원본유지)] model.language_model.layers.27.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
126
-> [양자화 X (원본유지)] model.language_model.layers.27.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
127
-> [양자화 X (원본유지)] model.language_model.layers.27.altup.router_norm.weight : 원본 형태 torch.Size([2048])
128
-> [양자화 X (원본유지)] model.language_model.layers.27.input_layernorm.weight : 원본 형태 torch.Size([2048])
129
-> [양자화 X (원본유지)] model.language_model.layers.27.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
130
-> [양자화 X (원본유지)] model.language_model.layers.27.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
131
-> [양자화 X (원본유지)] model.language_model.layers.27.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
132
-> [양자화 O] model.language_model.layers.27.mlp.down_proj.weight : 원본 형태 (2048, 8192)
133
-> [양자화 O] model.language_model.layers.27.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
134
-> [양자화 O] model.language_model.layers.27.mlp.up_proj.weight : 원본 형태 (8192, 2048)
135
-> [양자화 X (원본유지)] model.language_model.layers.27.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
136
-> [양자화 X (원본유지)] model.language_model.layers.27.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
137
-> [양자화 X (원본유지)] model.language_model.layers.27.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
138
-> [양자화 X (원본유지)] model.language_model.layers.27.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
139
-> [양자화 X (원본유지)] model.language_model.layers.27.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
140
-> [양자화 X (원본유지)] model.language_model.layers.27.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
141
-> [양자화 X (원본유지)] model.language_model.layers.27.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
142
-> [양자화 O] model.language_model.layers.27.self_attn.k_proj.weight : 원본 형태 (512, 2048)
143
-> [양자화 O] model.language_model.layers.27.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
144
-> [양자화 X (원본유지)] model.language_model.layers.27.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
145
-> [양자화 O] model.language_model.layers.27.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
146
-> [양자화 O] model.language_model.layers.27.self_attn.v_proj.weight : 원본 형태 (512, 2048)
147
-> [양자화 X (원본유지)] model.language_model.layers.28.altup.correct_output_scale : 원본 형태 torch.Size([2048])
148
-> [양자화 X (원본유지)] model.language_model.layers.28.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
149
-> [양자화 X (원본유지)] model.language_model.layers.28.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
150
-> [양자화 X (원본유지)] model.language_model.layers.28.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
151
-> [양자화 X (원본유지)] model.language_model.layers.28.altup.router_norm.weight : 원본 형태 torch.Size([2048])
152
-> [양자화 X (원본유지)] model.language_model.layers.28.input_layernorm.weight : 원본 형태 torch.Size([2048])
153
-> [양자화 X (원본유지)] model.language_model.layers.28.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
154
-> [양자화 X (원본유지)] model.language_model.layers.28.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
155
-> [양자화 X (원본유지)] model.language_model.layers.28.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
156
-> [양자화 O] model.language_model.layers.28.mlp.down_proj.weight : 원본 형태 (2048, 8192)
157
-> [양자화 O] model.language_model.layers.28.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
158
-> [양자화 O] model.language_model.layers.28.mlp.up_proj.weight : 원본 형태 (8192, 2048)
159
-> [양자화 X (원본유지)] model.language_model.layers.28.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
160
-> [양자화 X (원본유지)] model.language_model.layers.28.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
161
-> [양자화 X (원본유지)] model.language_model.layers.28.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
162
-> [양자화 X (원본유지)] model.language_model.layers.28.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
163
-> [양자화 X (원본유지)] model.language_model.layers.28.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
164
-> [양자화 X (원본유지)] model.language_model.layers.28.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
165
-> [양자화 X (원본유지)] model.language_model.layers.28.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
166
-> [양자화 O] model.language_model.layers.28.self_attn.k_proj.weight : 원본 형태 (512, 2048)
167
-> [양자화 O] model.language_model.layers.28.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
168
-> [양자화 X (원본유지)] model.language_model.layers.28.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
169
-> [양자화 O] model.language_model.layers.28.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
170
-> [양자화 O] model.language_model.layers.28.self_attn.v_proj.weight : 원본 형태 (512, 2048)
171
-> [양자화 X (원본유지)] model.language_model.layers.29.altup.correct_output_scale : 원본 형태 torch.Size([2048])
172
-> [양자화 X (원본유지)] model.language_model.layers.29.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
173
-> [양자화 X (원본유지)] model.language_model.layers.29.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
174
-> [양자화 X (원본유지)] model.language_model.layers.29.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
175
-> [양자화 X (원본유지)] model.language_model.layers.29.altup.router_norm.weight : 원본 형태 torch.Size([2048])
176
-> [양자화 X (원본유지)] model.language_model.layers.29.input_layernorm.weight : 원본 형태 torch.Size([2048])
177
-> [양자화 X (원본유지)] model.language_model.layers.29.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
178
-> [양자화 X (원본유지)] model.language_model.layers.29.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
179
-> [양자화 X (원본유지)] model.language_model.layers.29.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
180
-> [양자화 O] model.language_model.layers.29.mlp.down_proj.weight : 원본 형태 (2048, 8192)
181
-> [양자화 O] model.language_model.layers.29.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
182
-> [양자화 O] model.language_model.layers.29.mlp.up_proj.weight : 원본 형태 (8192, 2048)
183
-> [양자화 X (원본유지)] model.language_model.layers.29.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
184
-> [양자화 X (원본유지)] model.language_model.layers.29.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
185
-> [양자화 X (원본유지)] model.language_model.layers.29.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
186
-> [양자화 X (원본유지)] model.language_model.layers.29.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
187
-> [양자화 X (원본유지)] model.language_model.layers.29.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
188
-> [양자화 X (원본유지)] model.language_model.layers.29.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
189
-> [양자화 X (원본유지)] model.language_model.layers.29.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
190
-> [양자화 O] model.language_model.layers.29.self_attn.k_proj.weight : 원본 형태 (512, 2048)
191
-> [양자화 O] model.language_model.layers.29.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
192
-> [양자화 X (원본유지)] model.language_model.layers.29.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
193
-> [양자화 O] model.language_model.layers.29.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
194
-> [양자화 O] model.language_model.layers.29.self_attn.v_proj.weight : 원본 형태 (512, 2048)
195
-> [양자화 X (원본유지)] model.language_model.norm.weight : 원본 형태 torch.Size([2048])
196
-> [양자화 X (원본유지)] model.language_model.per_layer_model_projection.weight : 원본 형태 torch.Size([7680, 2048])
197
-> [양자화 X (원본유지)] model.language_model.per_layer_projection_norm.weight : 원본 형태 torch.Size([256])
198
저장 완료: Master/gemma3NE4B_INT4_Q/quantized_model-00002-of-00003.safetensors
199
200
[model-00003-of-00003.safetensors] 처리 중...
201
-> [양자화 O] model.audio_tower.conformer.10.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
202
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
203
-> [양자화 O] model.audio_tower.conformer.10.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
204
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
205
-> [양자화 O] model.audio_tower.conformer.10.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
206
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
207
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.attention.post_norm.weight : 원본 형태 torch.Size([1536])
208
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
209
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
210
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
211
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
212
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
213
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
214
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
215
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
216
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
217
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
218
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
219
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
220
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
221
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
222
-> [양자화 X (원본유지)] model.audio_tower.conformer.10.norm.weight : 원본 형태 torch.Size([1536])
223
-> [양자화 O] model.audio_tower.conformer.11.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
224
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
225
-> [양자화 O] model.audio_tower.conformer.11.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
226
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
227
-> [양자화 O] model.audio_tower.conformer.11.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
228
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
229
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.attention.post_norm.weight : 원본 형태 torch.Size([1536])
230
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
231
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
232
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
233
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
234
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
235
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
236
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
237
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
238
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
239
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
240
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
241
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
242
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
243
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
244
-> [양자화 X (원본유지)] model.audio_tower.conformer.11.norm.weight : 원본 형태 torch.Size([1536])
245
-> [양자화 O] model.audio_tower.conformer.4.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
246
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
247
-> [양자화 O] model.audio_tower.conformer.4.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
248
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
249
-> [양자화 O] model.audio_tower.conformer.4.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
250
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
251
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.attention.post_norm.weight : 원본 형태 torch.Size([1536])
252
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
253
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
254
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
255
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
256
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
257
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
258
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
259
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
260
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
261
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
262
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
263
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
264
-> [양자화 X (원본유지)] model.audio_tower.conformer.4.norm.weight : 원본 형태 torch.Size([1536])
265
-> [양자화 O] model.audio_tower.conformer.5.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
266
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
267
-> [양자화 O] model.audio_tower.conformer.5.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
268
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
269
-> [양자화 O] model.audio_tower.conformer.5.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
270
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
271
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.attention.post_norm.weight : 원본 형태 torch.Size([1536])
272
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
273
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
274
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
275
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
276
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
277
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
278
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
279
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
280
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
281
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
282
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
283
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
284
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
285
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
286
-> [양자화 X (원본유지)] model.audio_tower.conformer.5.norm.weight : 원본 형태 torch.Size([1536])
287
-> [양자화 O] model.audio_tower.conformer.6.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
288
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
289
-> [양자화 O] model.audio_tower.conformer.6.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
290
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
291
-> [양자화 O] model.audio_tower.conformer.6.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
292
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
293
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.attention.post_norm.weight : 원본 형태 torch.Size([1536])
294
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
295
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
296
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
297
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
298
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
299
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
300
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
301
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
302
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
303
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
304
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
305
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
306
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
307
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
308
-> [양자화 X (원본유지)] model.audio_tower.conformer.6.norm.weight : 원본 형태 torch.Size([1536])
309
-> [양자화 O] model.audio_tower.conformer.7.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
310
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
311
-> [양자화 O] model.audio_tower.conformer.7.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
312
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
313
-> [양자화 O] model.audio_tower.conformer.7.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
314
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
315
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.attention.post_norm.weight : 원본 형태 torch.Size([1536])
316
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
317
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
318
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
319
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
320
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
321
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
322
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
323
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
324
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
325
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
326
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
327
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
328
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
329
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
330
-> [양자화 X (원본유지)] model.audio_tower.conformer.7.norm.weight : 원본 형태 torch.Size([1536])
331
-> [양자화 O] model.audio_tower.conformer.8.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
332
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
333
-> [양자화 O] model.audio_tower.conformer.8.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
334
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
335
-> [양자화 O] model.audio_tower.conformer.8.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
336
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
337
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.attention.post_norm.weight : 원본 형태 torch.Size([1536])
338
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
339
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
340
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
341
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
342
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
343
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
344
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
345
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
346
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
347
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
348
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
349
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
350
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
351
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
352
-> [양자화 X (원본유지)] model.audio_tower.conformer.8.norm.weight : 원본 형태 torch.Size([1536])
353
-> [양자화 O] model.audio_tower.conformer.9.attention.attn.k_proj.weight : 원본 형태 (1536, 1536)
354
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.attention.attn.per_dim_scale : 원본 형태 torch.Size([192])
355
-> [양자화 O] model.audio_tower.conformer.9.attention.attn.q_proj.weight : 원본 형태 (1536, 1536)
356
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.attention.attn.relative_position_embedding.pos_proj.weight : 원본 형태 torch.Size([1536, 1536])
357
-> [양자화 O] model.audio_tower.conformer.9.attention.attn.v_proj.weight : 원본 형태 (1536, 1536)
358
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.attention.post.weight : 원본 형태 torch.Size([1536, 1536])
359
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.attention.post_norm.weight : 원본 형태 torch.Size([1536])
360
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.attention.pre_attn_norm.weight : 원본 형태 torch.Size([1536])
361
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_end.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
362
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_end.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
363
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_end.post_layer_norm.weight : 원본 형태 torch.Size([1536])
364
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_end.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
365
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_start.ffw_layer_1.weight : 원본 형태 torch.Size([6144, 1536])
366
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_start.ffw_layer_2.weight : 원본 형태 torch.Size([1536, 6144])
367
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_start.post_layer_norm.weight : 원본 형태 torch.Size([1536])
368
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.ffw_layer_start.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
369
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.lconv1d.conv_norm.weight : 원본 형태 torch.Size([1536])
370
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.lconv1d.depthwise_conv1d.weight : 원본 형태 torch.Size([1536, 1, 5])
371
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.lconv1d.linear_end.weight : 원본 형태 torch.Size([1536, 1536])
372
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.lconv1d.linear_start.weight : 원본 형태 torch.Size([3072, 1536])
373
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.lconv1d.pre_layer_norm.weight : 원본 형태 torch.Size([1536])
374
-> [양자화 X (원본유지)] model.audio_tower.conformer.9.norm.weight : 원본 형태 torch.Size([1536])
375
-> [양자화 X (원본유지)] model.embed_audio.embedding.weight : 원본 형태 torch.Size([128, 1536])
376
-> [양자화 X (원본유지)] model.embed_audio.embedding_projection.weight : 원본 형태 torch.Size([2048, 1536])
377
-> [양자화 X (원본유지)] model.embed_audio.hard_embedding_norm.weight : 원본 형태 torch.Size([1536])
378
-> [양자화 X (원본유지)] model.embed_audio.soft_embedding_norm.weight : 원본 형태 torch.Size([1536])
379
-> [양자화 X (원본유지)] model.embed_vision.embedding.weight : 원본 형태 torch.Size([128, 2048])
380
-> [양자화 X (원본유지)] model.embed_vision.embedding_projection.weight : 원본 형태 torch.Size([2048, 2048])
381
-> [양자화 X (원본유지)] model.embed_vision.hard_embedding_norm.weight : 원본 형태 torch.Size([2048])
382
-> [양자화 X (원본유지)] model.embed_vision.soft_embedding_norm.weight : 원본 형태 torch.Size([2048])
383
저장 완료: Master/gemma3NE4B_INT4_Q/quantized_model-00003-of-00003.safetensors
384
385
[model-00001-of-00003.safetensors] 처리 중...
386
-> [양자화 X (원본유지)] model.language_model.embed_tokens.weight : 원본 형태 torch.Size([262400, 2048])
387
-> [양자화 X (원본유지)] model.language_model.layers.0.altup.correct_output_scale : 원본 형태 torch.Size([2048])
388
-> [양자화 X (원본유지)] model.language_model.layers.0.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
389
-> [양자화 X (원본유지)] model.language_model.layers.0.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
390
-> [양자화 X (원본유지)] model.language_model.layers.0.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
391
-> [양자화 X (원본유지)] model.language_model.layers.0.altup.router_norm.weight : 원본 형태 torch.Size([2048])
392
-> [양자화 X (원본유지)] model.language_model.layers.0.input_layernorm.weight : 원본 형태 torch.Size([2048])
393
-> [양자화 X (원본유지)] model.language_model.layers.0.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
394
-> [양자화 X (원본유지)] model.language_model.layers.0.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
395
-> [양자화 X (원본유지)] model.language_model.layers.0.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
396
-> [양자화 O] model.language_model.layers.0.mlp.down_proj.weight : 원본 형태 (2048, 8192)
397
-> [양자화 O] model.language_model.layers.0.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
398
-> [양자화 O] model.language_model.layers.0.mlp.up_proj.weight : 원본 형태 (8192, 2048)
399
-> [양자화 X (원본유지)] model.language_model.layers.0.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
400
-> [양자화 X (원본유지)] model.language_model.layers.0.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
401
-> [양자화 X (원본유지)] model.language_model.layers.0.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
402
-> [양자화 X (원본유지)] model.language_model.layers.0.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
403
-> [양자화 X (원본유지)] model.language_model.layers.0.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
404
-> [양자화 X (원본유지)] model.language_model.layers.0.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
405
-> [양자화 X (원본유지)] model.language_model.layers.0.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
406
-> [양자화 O] model.language_model.layers.0.self_attn.k_proj.weight : 원본 형태 (512, 2048)
407
-> [양자화 O] model.language_model.layers.0.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
408
-> [양자화 X (원본유지)] model.language_model.layers.0.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
409
-> [양자화 O] model.language_model.layers.0.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
410
-> [양자화 O] model.language_model.layers.0.self_attn.v_proj.weight : 원본 형태 (512, 2048)
411
-> [양자화 X (원본유지)] model.language_model.layers.1.altup.correct_output_scale : 원본 형태 torch.Size([2048])
412
-> [양자화 X (원본유지)] model.language_model.layers.1.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
413
-> [양자화 X (원본유지)] model.language_model.layers.1.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
414
-> [양자화 X (원본유지)] model.language_model.layers.1.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
415
-> [양자화 X (원본유지)] model.language_model.layers.1.altup.router_norm.weight : 원본 형태 torch.Size([2048])
416
-> [양자화 X (원본유지)] model.language_model.layers.1.input_layernorm.weight : 원본 형태 torch.Size([2048])
417
-> [양자화 X (원본유지)] model.language_model.layers.1.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
418
-> [양자화 X (원본유지)] model.language_model.layers.1.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
419
-> [양자화 X (원본유지)] model.language_model.layers.1.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
420
-> [양자화 O] model.language_model.layers.1.mlp.down_proj.weight : 원본 형태 (2048, 8192)
421
-> [양자화 O] model.language_model.layers.1.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
422
-> [양자화 O] model.language_model.layers.1.mlp.up_proj.weight : 원본 형태 (8192, 2048)
423
-> [양자화 X (원본유지)] model.language_model.layers.1.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
424
-> [양자화 X (원본유지)] model.language_model.layers.1.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
425
-> [양자화 X (원본유지)] model.language_model.layers.1.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
426
-> [양자화 X (원본유지)] model.language_model.layers.1.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
427
-> [양자화 X (원본유지)] model.language_model.layers.1.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
428
-> [양자화 X (원본유지)] model.language_model.layers.1.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
429
-> [양자화 X (원본유지)] model.language_model.layers.1.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
430
-> [양자화 O] model.language_model.layers.1.self_attn.k_proj.weight : 원본 형태 (512, 2048)
431
-> [양자화 O] model.language_model.layers.1.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
432
-> [양자화 X (원본유지)] model.language_model.layers.1.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
433
-> [양자화 O] model.language_model.layers.1.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
434
-> [양자화 O] model.language_model.layers.1.self_attn.v_proj.weight : 원본 형태 (512, 2048)
435
-> [양자화 X (원본유지)] model.language_model.layers.10.altup.correct_output_scale : 원본 형태 torch.Size([2048])
436
-> [양자화 X (원본유지)] model.language_model.layers.10.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
437
-> [양자화 X (원본유지)] model.language_model.layers.10.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
438
-> [양자화 X (원본유지)] model.language_model.layers.10.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
439
-> [양자화 X (원본유지)] model.language_model.layers.10.altup.router_norm.weight : 원본 형태 torch.Size([2048])
440
-> [양자화 X (원본유지)] model.language_model.layers.10.input_layernorm.weight : 원본 형태 torch.Size([2048])
441
-> [양자화 X (원본유지)] model.language_model.layers.10.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
442
-> [양자화 X (원본유지)] model.language_model.layers.10.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
443
-> [양자화 X (원본유지)] model.language_model.layers.10.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
444
-> [양자화 O] model.language_model.layers.10.mlp.down_proj.weight : 원본 형태 (2048, 8192)
445
-> [양자화 O] model.language_model.layers.10.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
446
-> [양자화 O] model.language_model.layers.10.mlp.up_proj.weight : 원본 형태 (8192, 2048)
447
-> [양자화 X (원본유지)] model.language_model.layers.10.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
448
-> [양자화 X (원본유지)] model.language_model.layers.10.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
449
-> [양자화 X (원본유지)] model.language_model.layers.10.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
450
-> [양자화 X (원본유지)] model.language_model.layers.10.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
451
-> [양자화 X (원본유지)] model.language_model.layers.10.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
452
-> [양자화 X (원본유지)] model.language_model.layers.10.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
453
-> [양자화 X (원본유지)] model.language_model.layers.10.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
454
-> [양자화 O] model.language_model.layers.10.self_attn.k_proj.weight : 원본 형태 (512, 2048)
455
-> [양자화 O] model.language_model.layers.10.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
456
-> [양자화 X (원본유지)] model.language_model.layers.10.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
457
-> [양자화 O] model.language_model.layers.10.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
458
-> [양자화 O] model.language_model.layers.10.self_attn.v_proj.weight : 원본 형태 (512, 2048)
459
-> [양자화 X (원본유지)] model.language_model.layers.11.altup.correct_output_scale : 원본 형태 torch.Size([2048])
460
-> [양자화 X (원본유지)] model.language_model.layers.11.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
461
-> [양자화 X (원본유지)] model.language_model.layers.11.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
462
-> [양자화 X (원본유지)] model.language_model.layers.11.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
463
-> [양자화 X (원본유지)] model.language_model.layers.11.altup.router_norm.weight : 원본 형태 torch.Size([2048])
464
-> [양자화 X (원본유지)] model.language_model.layers.11.input_layernorm.weight : 원본 형태 torch.Size([2048])
465
-> [양자화 X (원본유지)] model.language_model.layers.11.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
466
-> [양자화 X (원본유지)] model.language_model.layers.11.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
467
-> [양자화 X (원본유지)] model.language_model.layers.11.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
468
-> [양자화 O] model.language_model.layers.11.mlp.down_proj.weight : 원본 형태 (2048, 8192)
469
-> [양자화 O] model.language_model.layers.11.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
470
-> [양자화 O] model.language_model.layers.11.mlp.up_proj.weight : 원본 형태 (8192, 2048)
471
-> [양자화 X (원본유지)] model.language_model.layers.11.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
472
-> [양자화 X (원본유지)] model.language_model.layers.11.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
473
-> [양자화 X (원본유지)] model.language_model.layers.11.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
474
-> [양자화 X (원본유지)] model.language_model.layers.11.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
475
-> [양자화 X (원본유지)] model.language_model.layers.11.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
476
-> [양자화 X (원본유지)] model.language_model.layers.11.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
477
-> [양자화 X (원본유지)] model.language_model.layers.11.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
478
-> [양자화 O] model.language_model.layers.11.self_attn.k_proj.weight : 원본 형태 (512, 2048)
479
-> [양자화 O] model.language_model.layers.11.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
480
-> [양자화 X (원본유지)] model.language_model.layers.11.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
481
-> [양자화 O] model.language_model.layers.11.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
482
-> [양자화 O] model.language_model.layers.11.self_attn.v_proj.weight : 원본 형태 (512, 2048)
483
-> [양자화 X (원본유지)] model.language_model.layers.12.altup.correct_output_scale : 원본 형태 torch.Size([2048])
484
-> [양자화 X (원본유지)] model.language_model.layers.12.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
485
-> [양자화 X (원본유지)] model.language_model.layers.12.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
486
-> [양자화 X (원본유지)] model.language_model.layers.12.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
487
-> [양자화 X (원본유지)] model.language_model.layers.12.altup.router_norm.weight : 원본 형태 torch.Size([2048])
488
-> [양자화 X (원본유지)] model.language_model.layers.12.input_layernorm.weight : 원본 형태 torch.Size([2048])
489
-> [양자화 X (원본유지)] model.language_model.layers.12.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
490
-> [양자화 X (원본유지)] model.language_model.layers.12.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
491
-> [양자화 X (원본유지)] model.language_model.layers.12.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
492
-> [양자화 O] model.language_model.layers.12.mlp.down_proj.weight : 원본 형태 (2048, 8192)
493
-> [양자화 O] model.language_model.layers.12.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
494
-> [양자화 O] model.language_model.layers.12.mlp.up_proj.weight : 원본 형태 (8192, 2048)
495
-> [양자화 X (원본유지)] model.language_model.layers.12.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
496
-> [양자화 X (원본유지)] model.language_model.layers.12.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
497
-> [양자화 X (원본유지)] model.language_model.layers.12.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
498
-> [양자화 X (원본유지)] model.language_model.layers.12.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
499
-> [양자화 X (원본유지)] model.language_model.layers.12.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
500
-> [양자화 X (원본유지)] model.language_model.layers.12.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
501
-> [양자화 X (원본유지)] model.language_model.layers.12.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
502
-> [양자화 O] model.language_model.layers.12.self_attn.k_proj.weight : 원본 형태 (512, 2048)
503
-> [양자화 O] model.language_model.layers.12.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
504
-> [양자화 X (원본유지)] model.language_model.layers.12.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
505
-> [양자화 O] model.language_model.layers.12.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
506
-> [양자화 O] model.language_model.layers.12.self_attn.v_proj.weight : 원본 형태 (512, 2048)
507
-> [양자화 X (원본유지)] model.language_model.layers.13.altup.correct_output_scale : 원본 형태 torch.Size([2048])
508
-> [양자화 X (원본유지)] model.language_model.layers.13.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
509
-> [양자화 X (원본유지)] model.language_model.layers.13.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
510
-> [양자화 X (원본유지)] model.language_model.layers.13.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
511
-> [양자화 X (원본유지)] model.language_model.layers.13.altup.router_norm.weight : 원본 형태 torch.Size([2048])
512
-> [양자화 X (원본유지)] model.language_model.layers.13.input_layernorm.weight : 원본 형태 torch.Size([2048])
513
-> [양자화 X (원본유지)] model.language_model.layers.13.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
514
-> [양자화 X (원본유지)] model.language_model.layers.13.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
515
-> [양자화 X (원본유지)] model.language_model.layers.13.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
516
-> [양자화 O] model.language_model.layers.13.mlp.down_proj.weight : 원본 형태 (2048, 8192)
517
-> [양자화 O] model.language_model.layers.13.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
518
-> [양자화 O] model.language_model.layers.13.mlp.up_proj.weight : 원본 형태 (8192, 2048)
519
-> [양자화 X (원본유지)] model.language_model.layers.13.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
520
-> [양자화 X (원본유지)] model.language_model.layers.13.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
521
-> [양자화 X (원본유지)] model.language_model.layers.13.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
522
-> [양자화 X (원본유지)] model.language_model.layers.13.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
523
-> [양자화 X (원본유지)] model.language_model.layers.13.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
524
-> [양자화 X (원본유지)] model.language_model.layers.13.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
525
-> [양자화 X (원본유지)] model.language_model.layers.13.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
526
-> [양자화 O] model.language_model.layers.13.self_attn.k_proj.weight : 원본 형태 (512, 2048)
527
-> [양자화 O] model.language_model.layers.13.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
528
-> [양자화 X (원본유지)] model.language_model.layers.13.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
529
-> [양자화 O] model.language_model.layers.13.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
530
-> [양자화 O] model.language_model.layers.13.self_attn.v_proj.weight : 원본 형태 (512, 2048)
531
-> [양자화 X (원본유지)] model.language_model.layers.14.altup.correct_output_scale : 원본 형태 torch.Size([2048])
532
-> [양자화 X (원본유지)] model.language_model.layers.14.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
533
-> [양자화 X (원본유지)] model.language_model.layers.14.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
534
-> [양자화 X (원본유지)] model.language_model.layers.14.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
535
-> [양자화 X (원본유지)] model.language_model.layers.14.altup.router_norm.weight : 원본 형태 torch.Size([2048])
536
-> [양자화 X (원본유지)] model.language_model.layers.14.input_layernorm.weight : 원본 형태 torch.Size([2048])
537
-> [양자화 X (원본유지)] model.language_model.layers.14.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
538
-> [양자화 X (원본유지)] model.language_model.layers.14.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
539
-> [양자화 X (원본유지)] model.language_model.layers.14.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
540
-> [양자화 O] model.language_model.layers.14.mlp.down_proj.weight : 원본 형태 (2048, 8192)
541
-> [양자화 O] model.language_model.layers.14.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
542
-> [양자화 O] model.language_model.layers.14.mlp.up_proj.weight : 원본 형태 (8192, 2048)
543
-> [양자화 X (원본유지)] model.language_model.layers.14.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
544
-> [양자화 X (원본유지)] model.language_model.layers.14.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
545
-> [양자화 X (원본유지)] model.language_model.layers.14.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
546
-> [양자화 X (원본유지)] model.language_model.layers.14.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
547
-> [양자화 X (원본유지)] model.language_model.layers.14.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
548
-> [양자화 X (원본유지)] model.language_model.layers.14.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
549
-> [양자화 X (원본유지)] model.language_model.layers.14.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
550
-> [양자화 O] model.language_model.layers.14.self_attn.k_proj.weight : 원본 형태 (512, 2048)
551
-> [양자화 O] model.language_model.layers.14.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
552
-> [양자화 X (원본유지)] model.language_model.layers.14.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
553
-> [양자화 O] model.language_model.layers.14.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
554
-> [양자화 O] model.language_model.layers.14.self_attn.v_proj.weight : 원본 형태 (512, 2048)
555
-> [양자화 X (원본유지)] model.language_model.layers.15.altup.correct_output_scale : 원본 형태 torch.Size([2048])
556
-> [양자화 X (원본유지)] model.language_model.layers.15.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
557
-> [양자화 X (원본유지)] model.language_model.layers.15.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
558
-> [양자화 X (원본유지)] model.language_model.layers.15.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
559
-> [양자화 X (원본유지)] model.language_model.layers.15.altup.router_norm.weight : 원본 형태 torch.Size([2048])
560
-> [양자화 X (원본유지)] model.language_model.layers.15.input_layernorm.weight : 원본 형태 torch.Size([2048])
561
-> [양자화 X (원본유지)] model.language_model.layers.15.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
562
-> [양자화 X (원본유지)] model.language_model.layers.15.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
563
-> [양자화 X (원본유지)] model.language_model.layers.15.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
564
-> [양자화 O] model.language_model.layers.15.mlp.down_proj.weight : 원본 형태 (2048, 8192)
565
-> [양자화 O] model.language_model.layers.15.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
566
-> [양자화 O] model.language_model.layers.15.mlp.up_proj.weight : 원본 형태 (8192, 2048)
567
-> [양자화 X (원본유지)] model.language_model.layers.15.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
568
-> [양자화 X (원본유지)] model.language_model.layers.15.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
569
-> [양자화 X (원본유지)] model.language_model.layers.15.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
570
-> [양자화 X (원본유지)] model.language_model.layers.15.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
571
-> [양자화 X (원본유지)] model.language_model.layers.15.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
572
-> [양자화 X (원본유지)] model.language_model.layers.15.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
573
-> [양자화 X (원본유지)] model.language_model.layers.15.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
574
-> [양자화 O] model.language_model.layers.15.self_attn.k_proj.weight : 원본 형태 (512, 2048)
575
-> [양자화 O] model.language_model.layers.15.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
576
-> [양자화 X (원본유지)] model.language_model.layers.15.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
577
-> [양자화 O] model.language_model.layers.15.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
578
-> [양자화 O] model.language_model.layers.15.self_attn.v_proj.weight : 원본 형태 (512, 2048)
579
-> [양자화 X (원본유지)] model.language_model.layers.16.altup.correct_output_scale : 원본 형태 torch.Size([2048])
580
-> [양자화 X (원본유지)] model.language_model.layers.16.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
581
-> [양자화 X (원본유지)] model.language_model.layers.16.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
582
-> [양자화 X (원본유지)] model.language_model.layers.16.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
583
-> [양자화 X (원본유지)] model.language_model.layers.16.altup.router_norm.weight : 원본 형태 torch.Size([2048])
584
-> [양자화 X (원본유지)] model.language_model.layers.16.input_layernorm.weight : 원본 형태 torch.Size([2048])
585
-> [양자화 X (원본유지)] model.language_model.layers.16.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
586
-> [양자화 X (원본유지)] model.language_model.layers.16.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
587
-> [양자화 X (원본유지)] model.language_model.layers.16.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
588
-> [양자화 O] model.language_model.layers.16.mlp.down_proj.weight : 원본 형태 (2048, 8192)
589
-> [양자화 O] model.language_model.layers.16.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
590
-> [양자화 O] model.language_model.layers.16.mlp.up_proj.weight : 원본 형태 (8192, 2048)
591
-> [양자화 X (원본유지)] model.language_model.layers.16.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
592
-> [양자화 X (원본유지)] model.language_model.layers.16.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
593
-> [양자화 X (원본유지)] model.language_model.layers.16.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
594
-> [양자화 X (원본유지)] model.language_model.layers.16.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
595
-> [양자화 X (원본유지)] model.language_model.layers.16.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
596
-> [양자화 X (원본유지)] model.language_model.layers.16.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
597
-> [양자화 X (원본유지)] model.language_model.layers.16.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
598
-> [양자화 O] model.language_model.layers.16.self_attn.k_proj.weight : 원본 형태 (512, 2048)
599
-> [양자화 O] model.language_model.layers.16.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
600
-> [양자화 X (원본유지)] model.language_model.layers.16.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
601
-> [양자화 O] model.language_model.layers.16.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
602
-> [양자화 O] model.language_model.layers.16.self_attn.v_proj.weight : 원본 형태 (512, 2048)
603
-> [양자화 X (원본유지)] model.language_model.layers.17.altup.correct_output_scale : 원본 형태 torch.Size([2048])
604
-> [양자화 X (원본유지)] model.language_model.layers.17.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
605
-> [양자화 X (원본유지)] model.language_model.layers.17.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
606
-> [양자화 X (원본유지)] model.language_model.layers.17.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
607
-> [양자화 X (원본유지)] model.language_model.layers.17.altup.router_norm.weight : 원본 형태 torch.Size([2048])
608
-> [양자화 X (원본유지)] model.language_model.layers.17.input_layernorm.weight : 원본 형태 torch.Size([2048])
609
-> [양자화 X (원본유지)] model.language_model.layers.17.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
610
-> [양자화 X (원본유지)] model.language_model.layers.17.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
611
-> [양자화 X (원본유지)] model.language_model.layers.17.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
612
-> [양자화 O] model.language_model.layers.17.mlp.down_proj.weight : 원본 형태 (2048, 8192)
613
-> [양자화 O] model.language_model.layers.17.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
614
-> [양자화 O] model.language_model.layers.17.mlp.up_proj.weight : 원본 형태 (8192, 2048)
615
-> [양자화 X (원본유지)] model.language_model.layers.17.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
616
-> [양자화 X (원본유지)] model.language_model.layers.17.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
617
-> [양자화 X (원본유지)] model.language_model.layers.17.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
618
-> [양자화 X (원본유지)] model.language_model.layers.17.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
619
-> [양자화 X (원본유지)] model.language_model.layers.17.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
620
-> [양자화 X (원본유지)] model.language_model.layers.17.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
621
-> [양자화 X (원본유지)] model.language_model.layers.17.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
622
-> [양자화 O] model.language_model.layers.17.self_attn.k_proj.weight : 원본 형태 (512, 2048)
623
-> [양자화 O] model.language_model.layers.17.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
624
-> [양자화 X (원본유지)] model.language_model.layers.17.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
625
-> [양자화 O] model.language_model.layers.17.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
626
-> [양자화 O] model.language_model.layers.17.self_attn.v_proj.weight : 원본 형태 (512, 2048)
627
-> [양자화 X (원본유지)] model.language_model.layers.18.altup.correct_output_scale : 원본 형태 torch.Size([2048])
628
-> [양자화 X (원본유지)] model.language_model.layers.18.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
629
-> [양자화 X (원본유지)] model.language_model.layers.18.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
630
-> [양자화 X (원본유지)] model.language_model.layers.18.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
631
-> [양자화 X (원본유지)] model.language_model.layers.18.altup.router_norm.weight : 원본 형태 torch.Size([2048])
632
-> [양자화 X (원본유지)] model.language_model.layers.18.input_layernorm.weight : 원본 형태 torch.Size([2048])
633
-> [양자화 X (원본유지)] model.language_model.layers.18.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
634
-> [양자화 X (원본유지)] model.language_model.layers.18.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
635
-> [양자화 X (원본유지)] model.language_model.layers.18.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
636
-> [양자화 O] model.language_model.layers.18.mlp.down_proj.weight : 원본 형태 (2048, 8192)
637
-> [양자화 O] model.language_model.layers.18.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
638
-> [양자화 O] model.language_model.layers.18.mlp.up_proj.weight : 원본 형태 (8192, 2048)
639
-> [양자화 X (원본유지)] model.language_model.layers.18.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
640
-> [양자화 X (원본유지)] model.language_model.layers.18.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
641
-> [양자화 X (원본유지)] model.language_model.layers.18.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
642
-> [양자화 X (원본유지)] model.language_model.layers.18.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
643
-> [양자화 X (원본유지)] model.language_model.layers.18.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
644
-> [양자화 X (원본유지)] model.language_model.layers.18.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
645
-> [양자화 X (원본유지)] model.language_model.layers.18.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
646
-> [양자화 O] model.language_model.layers.18.self_attn.k_proj.weight : 원본 형태 (512, 2048)
647
-> [양자화 O] model.language_model.layers.18.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
648
-> [양자화 X (원본유지)] model.language_model.layers.18.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
649
-> [양자화 O] model.language_model.layers.18.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
650
-> [양자화 O] model.language_model.layers.18.self_attn.v_proj.weight : 원본 형태 (512, 2048)
651
-> [양자화 X (원본유지)] model.language_model.layers.19.altup.correct_output_scale : 원본 형태 torch.Size([2048])
652
-> [양자화 X (원본유지)] model.language_model.layers.19.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
653
-> [양자화 X (원본유지)] model.language_model.layers.19.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
654
-> [양자화 X (원본유지)] model.language_model.layers.19.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
655
-> [양자화 X (원본유지)] model.language_model.layers.19.altup.router_norm.weight : 원본 형태 torch.Size([2048])
656
-> [양자화 X (원본유지)] model.language_model.layers.19.input_layernorm.weight : 원본 형태 torch.Size([2048])
657
-> [양자화 X (원본유지)] model.language_model.layers.19.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
658
-> [양자화 X (원본유지)] model.language_model.layers.19.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
659
-> [양자화 X (원본유지)] model.language_model.layers.19.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
660
-> [양자화 O] model.language_model.layers.19.mlp.down_proj.weight : 원본 형태 (2048, 8192)
661
-> [양자화 O] model.language_model.layers.19.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
662
-> [양자화 O] model.language_model.layers.19.mlp.up_proj.weight : 원본 형태 (8192, 2048)
663
-> [양자화 X (원본유지)] model.language_model.layers.19.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
664
-> [양자화 X (원본유지)] model.language_model.layers.19.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
665
-> [양자화 X (원본유지)] model.language_model.layers.19.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
666
-> [양자화 X (원본유지)] model.language_model.layers.19.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
667
-> [양자화 X (원본유지)] model.language_model.layers.19.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
668
-> [양자화 X (원본유지)] model.language_model.layers.19.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
669
-> [양자화 X (원본유지)] model.language_model.layers.19.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
670
-> [양자화 O] model.language_model.layers.19.self_attn.k_proj.weight : 원본 형태 (512, 2048)
671
-> [양자화 O] model.language_model.layers.19.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
672
-> [양자화 X (원본유지)] model.language_model.layers.19.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
673
-> [양자화 O] model.language_model.layers.19.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
674
-> [양자화 O] model.language_model.layers.19.self_attn.v_proj.weight : 원본 형태 (512, 2048)
675
-> [양자화 X (원본유지)] model.language_model.layers.2.altup.correct_output_scale : 원본 형태 torch.Size([2048])
676
-> [양자화 X (원본유지)] model.language_model.layers.2.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
677
-> [양자화 X (원본유지)] model.language_model.layers.2.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
678
-> [양자화 X (원본유지)] model.language_model.layers.2.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
679
-> [양자화 X (원본유지)] model.language_model.layers.2.altup.router_norm.weight : 원본 형태 torch.Size([2048])
680
-> [양자화 X (원본유지)] model.language_model.layers.2.input_layernorm.weight : 원본 형태 torch.Size([2048])
681
-> [양자화 X (원본유지)] model.language_model.layers.2.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
682
-> [양자화 X (원본유지)] model.language_model.layers.2.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
683
-> [양자화 X (원본유지)] model.language_model.layers.2.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
684
-> [양자화 O] model.language_model.layers.2.mlp.down_proj.weight : 원본 형태 (2048, 8192)
685
-> [양자화 O] model.language_model.layers.2.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
686
-> [양자화 O] model.language_model.layers.2.mlp.up_proj.weight : 원본 형태 (8192, 2048)
687
-> [양자화 X (원본유지)] model.language_model.layers.2.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
688
-> [양자화 X (원본유지)] model.language_model.layers.2.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
689
-> [양자화 X (원본유지)] model.language_model.layers.2.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
690
-> [양자화 X (원본유지)] model.language_model.layers.2.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
691
-> [양자화 X (원본유지)] model.language_model.layers.2.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
692
-> [양자화 X (원본유지)] model.language_model.layers.2.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
693
-> [양자화 X (원본유지)] model.language_model.layers.2.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
694
-> [양자화 O] model.language_model.layers.2.self_attn.k_proj.weight : 원본 형태 (512, 2048)
695
-> [양자화 O] model.language_model.layers.2.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
696
-> [양자화 X (원본유지)] model.language_model.layers.2.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
697
-> [양자화 O] model.language_model.layers.2.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
698
-> [양자화 O] model.language_model.layers.2.self_attn.v_proj.weight : 원본 형태 (512, 2048)
699
-> [양자화 X (원본유지)] model.language_model.layers.20.altup.correct_output_scale : 원본 형태 torch.Size([2048])
700
-> [양자화 X (원본유지)] model.language_model.layers.20.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
701
-> [양자화 X (원본유지)] model.language_model.layers.20.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
702
-> [양자화 X (원본유지)] model.language_model.layers.20.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
703
-> [양자화 X (원본유지)] model.language_model.layers.20.altup.router_norm.weight : 원본 형태 torch.Size([2048])
704
-> [양자화 X (원본유지)] model.language_model.layers.20.input_layernorm.weight : 원본 형태 torch.Size([2048])
705
-> [양자화 X (원본유지)] model.language_model.layers.20.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
706
-> [양자화 X (원본유지)] model.language_model.layers.20.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
707
-> [양자화 X (원본유지)] model.language_model.layers.20.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
708
-> [양자화 O] model.language_model.layers.20.mlp.down_proj.weight : 원본 형태 (2048, 8192)
709
-> [양자화 O] model.language_model.layers.20.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
710
-> [양자화 O] model.language_model.layers.20.mlp.up_proj.weight : 원본 형태 (8192, 2048)
711
-> [양자화 X (원본유지)] model.language_model.layers.20.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
712
-> [양자화 X (원본유지)] model.language_model.layers.20.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
713
-> [양자화 X (원본유지)] model.language_model.layers.20.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
714
-> [양자화 X (원본유지)] model.language_model.layers.20.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
715
-> [양자화 X (원본유지)] model.language_model.layers.20.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
716
-> [양자화 X (원본유지)] model.language_model.layers.20.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
717
-> [양자화 X (원본유지)] model.language_model.layers.20.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
718
-> [양자화 O] model.language_model.layers.20.self_attn.k_proj.weight : 원본 형태 (512, 2048)
719
-> [양자화 O] model.language_model.layers.20.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
720
-> [양자화 X (원본유지)] model.language_model.layers.20.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
721
-> [양자화 O] model.language_model.layers.20.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
722
-> [양자화 O] model.language_model.layers.20.self_attn.v_proj.weight : 원본 형태 (512, 2048)
723
-> [양자화 X (원본유지)] model.language_model.layers.21.altup.correct_output_scale : 원본 형태 torch.Size([2048])
724
-> [양자화 X (원본유지)] model.language_model.layers.21.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
725
-> [양자화 X (원본유지)] model.language_model.layers.21.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
726
-> [양자화 X (원본유지)] model.language_model.layers.21.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
727
-> [양자화 X (원본유지)] model.language_model.layers.21.altup.router_norm.weight : 원본 형태 torch.Size([2048])
728
-> [양자화 X (원본유지)] model.language_model.layers.21.input_layernorm.weight : 원본 형태 torch.Size([2048])
729
-> [양자화 X (원본유지)] model.language_model.layers.21.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
730
-> [양자화 X (원본유지)] model.language_model.layers.21.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
731
-> [양자화 X (원본유지)] model.language_model.layers.21.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
732
-> [양자화 O] model.language_model.layers.21.mlp.down_proj.weight : 원본 형태 (2048, 8192)
733
-> [양자화 O] model.language_model.layers.21.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
734
-> [양자화 O] model.language_model.layers.21.mlp.up_proj.weight : 원본 형태 (8192, 2048)
735
-> [양자화 X (원본유지)] model.language_model.layers.21.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
736
-> [양자화 X (원본유지)] model.language_model.layers.21.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
737
-> [양자화 X (원본유지)] model.language_model.layers.21.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
738
-> [양자화 X (원본유지)] model.language_model.layers.21.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
739
-> [양자화 X (원본유지)] model.language_model.layers.21.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
740
-> [양자화 X (원본유지)] model.language_model.layers.21.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
741
-> [양자화 X (원본유지)] model.language_model.layers.21.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
742
-> [양자화 O] model.language_model.layers.21.self_attn.k_proj.weight : 원본 형태 (512, 2048)
743
-> [양자화 O] model.language_model.layers.21.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
744
-> [양자화 X (원본유지)] model.language_model.layers.21.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
745
-> [양자화 O] model.language_model.layers.21.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
746
-> [양자화 O] model.language_model.layers.21.self_attn.v_proj.weight : 원본 형태 (512, 2048)
747
-> [양자화 X (원본유지)] model.language_model.layers.22.altup.correct_output_scale : 원본 형태 torch.Size([2048])
748
-> [양자화 X (원본유지)] model.language_model.layers.22.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
749
-> [양자화 X (원본유지)] model.language_model.layers.22.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
750
-> [양자화 X (원본유지)] model.language_model.layers.22.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
751
-> [양자화 X (원본유지)] model.language_model.layers.22.altup.router_norm.weight : 원본 형태 torch.Size([2048])
752
-> [양자화 X (원본유지)] model.language_model.layers.22.input_layernorm.weight : 원본 형태 torch.Size([2048])
753
-> [양자화 X (원본유지)] model.language_model.layers.22.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
754
-> [양자화 X (원본유지)] model.language_model.layers.22.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
755
-> [양자화 X (원본유지)] model.language_model.layers.22.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
756
-> [양자화 O] model.language_model.layers.22.mlp.down_proj.weight : 원본 형태 (2048, 8192)
757
-> [양자화 O] model.language_model.layers.22.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
758
-> [양자화 O] model.language_model.layers.22.mlp.up_proj.weight : 원본 형태 (8192, 2048)
759
-> [양자화 X (원본유지)] model.language_model.layers.22.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
760
-> [양자화 X (원본유지)] model.language_model.layers.22.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
761
-> [양자화 X (원본유지)] model.language_model.layers.22.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
762
-> [양자화 X (원본유지)] model.language_model.layers.22.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
763
-> [양자화 X (원본유지)] model.language_model.layers.22.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
764
-> [양자화 X (원본유지)] model.language_model.layers.22.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
765
-> [양자화 X (원본유지)] model.language_model.layers.22.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
766
-> [양자화 O] model.language_model.layers.22.self_attn.k_proj.weight : 원본 형태 (512, 2048)
767
-> [양자화 O] model.language_model.layers.22.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
768
-> [양자화 X (원본유지)] model.language_model.layers.22.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
769
-> [양자화 O] model.language_model.layers.22.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
770
-> [양자화 O] model.language_model.layers.22.self_attn.v_proj.weight : 원본 형태 (512, 2048)
771
-> [양자화 X (원본유지)] model.language_model.layers.23.altup.correct_output_scale : 원본 형태 torch.Size([2048])
772
-> [양자화 X (원본유지)] model.language_model.layers.23.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
773
-> [양자화 X (원본유지)] model.language_model.layers.23.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
774
-> [양자화 X (원본유지)] model.language_model.layers.23.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
775
-> [양자화 X (원본유지)] model.language_model.layers.23.altup.router_norm.weight : 원본 형태 torch.Size([2048])
776
-> [양자화 X (원본유지)] model.language_model.layers.23.input_layernorm.weight : 원본 형태 torch.Size([2048])
777
-> [양자화 X (원본유지)] model.language_model.layers.23.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
778
-> [양자화 X (원본유지)] model.language_model.layers.23.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
779
-> [양자화 X (원본유지)] model.language_model.layers.23.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
780
-> [양자화 O] model.language_model.layers.23.mlp.down_proj.weight : 원본 형태 (2048, 8192)
781
-> [양자화 O] model.language_model.layers.23.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
782
-> [양자화 O] model.language_model.layers.23.mlp.up_proj.weight : 원본 형태 (8192, 2048)
783
-> [양자화 X (원본유지)] model.language_model.layers.23.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
784
-> [양자화 X (원본유지)] model.language_model.layers.23.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
785
-> [양자화 X (원본유지)] model.language_model.layers.23.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
786
-> [양자화 X (원본유지)] model.language_model.layers.23.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
787
-> [양자화 X (원본유지)] model.language_model.layers.23.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
788
-> [양자화 X (원본유지)] model.language_model.layers.23.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
789
-> [양자화 X (원본유지)] model.language_model.layers.23.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
790
-> [양자화 O] model.language_model.layers.23.self_attn.k_proj.weight : 원본 형태 (512, 2048)
791
-> [양자화 O] model.language_model.layers.23.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
792
-> [양자화 X (원본유지)] model.language_model.layers.23.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
793
-> [양자화 O] model.language_model.layers.23.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
794
-> [양자화 O] model.language_model.layers.23.self_attn.v_proj.weight : 원본 형태 (512, 2048)
795
-> [양자화 X (원본유지)] model.language_model.layers.24.altup.correct_output_scale : 원본 형태 torch.Size([2048])
796
-> [양자화 X (원본유지)] model.language_model.layers.24.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
797
-> [양자화 X (원본유지)] model.language_model.layers.24.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
798
-> [양자화 X (원본유지)] model.language_model.layers.24.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
799
-> [양자화 X (원본유지)] model.language_model.layers.24.altup.router_norm.weight : 원본 형태 torch.Size([2048])
800
-> [양자화 X (원본유지)] model.language_model.layers.24.input_layernorm.weight : 원본 형태 torch.Size([2048])
801
-> [양자화 X (원본유지)] model.language_model.layers.24.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
802
-> [양자화 X (원본유지)] model.language_model.layers.24.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
803
-> [양자화 X (원본유지)] model.language_model.layers.24.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
804
-> [양자화 O] model.language_model.layers.24.mlp.down_proj.weight : 원본 형태 (2048, 8192)
805
-> [양자화 O] model.language_model.layers.24.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
806
-> [양자화 O] model.language_model.layers.24.mlp.up_proj.weight : 원본 형태 (8192, 2048)
807
-> [양자화 X (원본유지)] model.language_model.layers.24.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
808
-> [양자화 X (원본유지)] model.language_model.layers.24.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
809
-> [양자화 X (원본유지)] model.language_model.layers.24.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
810
-> [양자화 X (원본유지)] model.language_model.layers.24.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
811
-> [양자화 X (원본유지)] model.language_model.layers.24.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
812
-> [양자화 X (원본유지)] model.language_model.layers.24.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
813
-> [양자화 X (원본유지)] model.language_model.layers.24.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
814
-> [양자화 O] model.language_model.layers.24.self_attn.k_proj.weight : 원본 형태 (512, 2048)
815
-> [양자화 O] model.language_model.layers.24.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
816
-> [양자화 X (원본유지)] model.language_model.layers.24.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
817
-> [양자화 O] model.language_model.layers.24.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
818
-> [양자화 O] model.language_model.layers.24.self_attn.v_proj.weight : 원본 형태 (512, 2048)
819
-> [양자화 X (원본유지)] model.language_model.layers.25.altup.correct_output_scale : 원본 형태 torch.Size([2048])
820
-> [양자화 X (원본유지)] model.language_model.layers.25.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
821
-> [양자화 X (원본유지)] model.language_model.layers.25.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
822
-> [양자화 X (원본유지)] model.language_model.layers.25.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
823
-> [양자화 X (원본유지)] model.language_model.layers.25.altup.router_norm.weight : 원본 형태 torch.Size([2048])
824
-> [양자화 X (원본유지)] model.language_model.layers.25.input_layernorm.weight : 원본 형태 torch.Size([2048])
825
-> [양자화 X (원본유지)] model.language_model.layers.25.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
826
-> [양자화 X (원본유지)] model.language_model.layers.25.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
827
-> [양자화 X (원본유지)] model.language_model.layers.25.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
828
-> [양자화 O] model.language_model.layers.25.mlp.down_proj.weight : 원본 형태 (2048, 8192)
829
-> [양자화 O] model.language_model.layers.25.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
830
-> [양자화 O] model.language_model.layers.25.mlp.up_proj.weight : 원본 형태 (8192, 2048)
831
-> [양자화 X (원본유지)] model.language_model.layers.25.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
832
-> [양자화 X (원본유지)] model.language_model.layers.25.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
833
-> [양자화 X (원본유지)] model.language_model.layers.25.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
834
-> [양자화 X (원본유지)] model.language_model.layers.25.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
835
-> [양자화 X (원본유지)] model.language_model.layers.25.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
836
-> [양자화 X (원본유지)] model.language_model.layers.25.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
837
-> [양자화 X (원본유지)] model.language_model.layers.25.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
838
-> [양자화 O] model.language_model.layers.25.self_attn.k_proj.weight : 원본 형태 (512, 2048)
839
-> [양자화 O] model.language_model.layers.25.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
840
-> [양자화 X (원본유지)] model.language_model.layers.25.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
841
-> [양자화 O] model.language_model.layers.25.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
842
-> [양자화 O] model.language_model.layers.25.self_attn.v_proj.weight : 원본 형태 (512, 2048)
843
-> [양자화 O] model.language_model.layers.26.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
844
-> [양자화 O] model.language_model.layers.26.mlp.up_proj.weight : 원본 형태 (8192, 2048)
845
-> [양자화 X (원본유지)] model.language_model.layers.26.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
846
-> [양자화 O] model.language_model.layers.26.self_attn.k_proj.weight : 원본 형태 (512, 2048)
847
-> [양자화 O] model.language_model.layers.26.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
848
-> [양자화 X (원본유지)] model.language_model.layers.26.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
849
-> [양자화 O] model.language_model.layers.26.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
850
-> [양자화 O] model.language_model.layers.26.self_attn.v_proj.weight : 원본 형태 (512, 2048)
851
-> [양자화 X (원본유지)] model.language_model.layers.3.altup.correct_output_scale : 원본 형태 torch.Size([2048])
852
-> [양자화 X (원본유지)] model.language_model.layers.3.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
853
-> [양자화 X (원본유지)] model.language_model.layers.3.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
854
-> [양자화 X (원본유지)] model.language_model.layers.3.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
855
-> [양자화 X (원본유지)] model.language_model.layers.3.altup.router_norm.weight : 원본 형태 torch.Size([2048])
856
-> [양자화 X (원본유지)] model.language_model.layers.3.input_layernorm.weight : 원본 형태 torch.Size([2048])
857
-> [양자화 X (원본유지)] model.language_model.layers.3.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
858
-> [양자화 X (원본유지)] model.language_model.layers.3.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
859
-> [양자화 X (원본유지)] model.language_model.layers.3.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
860
-> [양자화 O] model.language_model.layers.3.mlp.down_proj.weight : 원본 형태 (2048, 8192)
861
-> [양자화 O] model.language_model.layers.3.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
862
-> [양자화 O] model.language_model.layers.3.mlp.up_proj.weight : 원본 형태 (8192, 2048)
863
-> [양자화 X (원본유지)] model.language_model.layers.3.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
864
-> [양자화 X (원본유지)] model.language_model.layers.3.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
865
-> [양자화 X (원본유지)] model.language_model.layers.3.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
866
-> [양자화 X (원본유지)] model.language_model.layers.3.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
867
-> [양자화 X (원본유지)] model.language_model.layers.3.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
868
-> [양자화 X (원본유지)] model.language_model.layers.3.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
869
-> [양자화 X (원본유지)] model.language_model.layers.3.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
870
-> [양자화 O] model.language_model.layers.3.self_attn.k_proj.weight : 원본 형태 (512, 2048)
871
-> [양자화 O] model.language_model.layers.3.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
872
-> [양자화 X (원본유지)] model.language_model.layers.3.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
873
-> [양자화 O] model.language_model.layers.3.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
874
-> [양자화 O] model.language_model.layers.3.self_attn.v_proj.weight : 원본 형태 (512, 2048)
875
-> [양자화 X (원본유지)] model.language_model.layers.4.altup.correct_output_scale : 원본 형태 torch.Size([2048])
876
-> [양자화 X (원본유지)] model.language_model.layers.4.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
877
-> [양자화 X (원본유지)] model.language_model.layers.4.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
878
-> [양자화 X (원본유지)] model.language_model.layers.4.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
879
-> [양자화 X (원본유지)] model.language_model.layers.4.altup.router_norm.weight : 원본 형태 torch.Size([2048])
880
-> [양자화 X (원본유지)] model.language_model.layers.4.input_layernorm.weight : 원본 형태 torch.Size([2048])
881
-> [양자화 X (원본유지)] model.language_model.layers.4.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
882
-> [양자화 X (원본유지)] model.language_model.layers.4.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
883
-> [양자화 X (원본유지)] model.language_model.layers.4.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
884
-> [양자화 O] model.language_model.layers.4.mlp.down_proj.weight : 원본 형태 (2048, 8192)
885
-> [양자화 O] model.language_model.layers.4.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
886
-> [양자화 O] model.language_model.layers.4.mlp.up_proj.weight : 원본 형태 (8192, 2048)
887
-> [양자화 X (원본유지)] model.language_model.layers.4.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
888
-> [양자화 X (원본유지)] model.language_model.layers.4.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
889
-> [양자화 X (원본유지)] model.language_model.layers.4.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
890
-> [양자화 X (원본유지)] model.language_model.layers.4.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
891
-> [양자화 X (원본유지)] model.language_model.layers.4.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
892
-> [양자화 X (원본유지)] model.language_model.layers.4.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
893
-> [양자화 X (원본유지)] model.language_model.layers.4.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
894
-> [양자화 O] model.language_model.layers.4.self_attn.k_proj.weight : 원본 형태 (512, 2048)
895
-> [양자화 O] model.language_model.layers.4.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
896
-> [양자화 X (원본유지)] model.language_model.layers.4.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
897
-> [양자화 O] model.language_model.layers.4.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
898
-> [양자화 O] model.language_model.layers.4.self_attn.v_proj.weight : 원본 형태 (512, 2048)
899
-> [양자화 X (원본유지)] model.language_model.layers.5.altup.correct_output_scale : 원본 형태 torch.Size([2048])
900
-> [양자화 X (원본유지)] model.language_model.layers.5.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
901
-> [양자화 X (원본유지)] model.language_model.layers.5.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
902
-> [양자화 X (원본유지)] model.language_model.layers.5.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
903
-> [양자화 X (원본유지)] model.language_model.layers.5.altup.router_norm.weight : 원본 형태 torch.Size([2048])
904
-> [양자화 X (원본유지)] model.language_model.layers.5.input_layernorm.weight : 원본 형태 torch.Size([2048])
905
-> [양자화 X (원본유지)] model.language_model.layers.5.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
906
-> [양자화 X (원본유지)] model.language_model.layers.5.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
907
-> [양자화 X (원본유지)] model.language_model.layers.5.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
908
-> [양자화 O] model.language_model.layers.5.mlp.down_proj.weight : 원본 형태 (2048, 8192)
909
-> [양자화 O] model.language_model.layers.5.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
910
-> [양자화 O] model.language_model.layers.5.mlp.up_proj.weight : 원본 형태 (8192, 2048)
911
-> [양자화 X (원본유지)] model.language_model.layers.5.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
912
-> [양자화 X (원본유지)] model.language_model.layers.5.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
913
-> [양자화 X (원본유지)] model.language_model.layers.5.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
914
-> [양자화 X (원본유지)] model.language_model.layers.5.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
915
-> [양자화 X (원본유지)] model.language_model.layers.5.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
916
-> [양자화 X (원본유지)] model.language_model.layers.5.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
917
-> [양자화 X (원본유지)] model.language_model.layers.5.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
918
-> [양자화 O] model.language_model.layers.5.self_attn.k_proj.weight : 원본 형태 (512, 2048)
919
-> [양자화 O] model.language_model.layers.5.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
920
-> [양자화 X (원본유지)] model.language_model.layers.5.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
921
-> [양자화 O] model.language_model.layers.5.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
922
-> [양자화 O] model.language_model.layers.5.self_attn.v_proj.weight : 원본 형태 (512, 2048)
923
-> [양자화 X (원본유지)] model.language_model.layers.6.altup.correct_output_scale : 원본 형태 torch.Size([2048])
924
-> [양자화 X (원본유지)] model.language_model.layers.6.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
925
-> [양자화 X (원본유지)] model.language_model.layers.6.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
926
-> [양자화 X (원본유지)] model.language_model.layers.6.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
927
-> [양자화 X (원본유지)] model.language_model.layers.6.altup.router_norm.weight : 원본 형태 torch.Size([2048])
928
-> [양자화 X (원본유지)] model.language_model.layers.6.input_layernorm.weight : 원본 형태 torch.Size([2048])
929
-> [양자화 X (원본유지)] model.language_model.layers.6.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
930
-> [양자화 X (원본유지)] model.language_model.layers.6.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
931
-> [양자화 X (원본유지)] model.language_model.layers.6.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
932
-> [양자화 O] model.language_model.layers.6.mlp.down_proj.weight : 원본 형태 (2048, 8192)
933
-> [양자화 O] model.language_model.layers.6.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
934
-> [양자화 O] model.language_model.layers.6.mlp.up_proj.weight : 원본 형태 (8192, 2048)
935
-> [양자화 X (원본유지)] model.language_model.layers.6.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
936
-> [양자화 X (원본유지)] model.language_model.layers.6.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
937
-> [양자화 X (원본유지)] model.language_model.layers.6.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
938
-> [양자화 X (원본유지)] model.language_model.layers.6.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
939
-> [양자화 X (원본유지)] model.language_model.layers.6.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
940
-> [양자화 X (원본유지)] model.language_model.layers.6.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
941
-> [양자화 X (원본유지)] model.language_model.layers.6.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
942
-> [양자화 O] model.language_model.layers.6.self_attn.k_proj.weight : 원본 형태 (512, 2048)
943
-> [양자화 O] model.language_model.layers.6.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
944
-> [양자화 X (원본유지)] model.language_model.layers.6.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
945
-> [양자화 O] model.language_model.layers.6.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
946
-> [양자화 O] model.language_model.layers.6.self_attn.v_proj.weight : 원본 형태 (512, 2048)
947
-> [양자화 X (원본유지)] model.language_model.layers.7.altup.correct_output_scale : 원본 형태 torch.Size([2048])
948
-> [양자화 X (원본유지)] model.language_model.layers.7.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
949
-> [양자화 X (원본유지)] model.language_model.layers.7.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
950
-> [양자화 X (원본유지)] model.language_model.layers.7.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
951
-> [양자화 X (원본유지)] model.language_model.layers.7.altup.router_norm.weight : 원본 형태 torch.Size([2048])
952
-> [양자화 X (원본유지)] model.language_model.layers.7.input_layernorm.weight : 원본 형태 torch.Size([2048])
953
-> [양자화 X (원본유지)] model.language_model.layers.7.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
954
-> [양자화 X (원본유지)] model.language_model.layers.7.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
955
-> [양자화 X (원본유지)] model.language_model.layers.7.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
956
-> [양자화 O] model.language_model.layers.7.mlp.down_proj.weight : 원본 형태 (2048, 8192)
957
-> [양자화 O] model.language_model.layers.7.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
958
-> [양자화 O] model.language_model.layers.7.mlp.up_proj.weight : 원본 형태 (8192, 2048)
959
-> [양자화 X (원본유지)] model.language_model.layers.7.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
960
-> [양자화 X (원본유지)] model.language_model.layers.7.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
961
-> [양자화 X (원본유지)] model.language_model.layers.7.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
962
-> [양자화 X (원본유지)] model.language_model.layers.7.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
963
-> [양자화 X (원본유지)] model.language_model.layers.7.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
964
-> [양자화 X (원본유지)] model.language_model.layers.7.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
965
-> [양자화 X (원본유지)] model.language_model.layers.7.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
966
-> [양자화 O] model.language_model.layers.7.self_attn.k_proj.weight : 원본 형태 (512, 2048)
967
-> [양자화 O] model.language_model.layers.7.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
968
-> [양자화 X (원본유지)] model.language_model.layers.7.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
969
-> [양자화 O] model.language_model.layers.7.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
970
-> [양자화 O] model.language_model.layers.7.self_attn.v_proj.weight : 원본 형태 (512, 2048)
971
-> [양자화 X (원본유지)] model.language_model.layers.8.altup.correct_output_scale : 원본 형태 torch.Size([2048])
972
-> [양자화 X (원본유지)] model.language_model.layers.8.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
973
-> [양자화 X (원본유지)] model.language_model.layers.8.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
974
-> [양자화 X (원본유지)] model.language_model.layers.8.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
975
-> [양자화 X (원본유지)] model.language_model.layers.8.altup.router_norm.weight : 원본 형태 torch.Size([2048])
976
-> [양자화 X (원본유지)] model.language_model.layers.8.input_layernorm.weight : 원본 형태 torch.Size([2048])
977
-> [양자화 X (원본유지)] model.language_model.layers.8.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
978
-> [양자화 X (원본유지)] model.language_model.layers.8.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
979
-> [양자화 X (원본유지)] model.language_model.layers.8.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
980
-> [양자화 O] model.language_model.layers.8.mlp.down_proj.weight : 원본 형태 (2048, 8192)
981
-> [양자화 O] model.language_model.layers.8.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
982
-> [양자화 O] model.language_model.layers.8.mlp.up_proj.weight : 원본 형태 (8192, 2048)
983
-> [양자화 X (원본유지)] model.language_model.layers.8.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
984
-> [양자화 X (원본유지)] model.language_model.layers.8.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
985
-> [양자화 X (원본유지)] model.language_model.layers.8.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
986
-> [양자화 X (원본유지)] model.language_model.layers.8.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
987
-> [양자화 X (원본유지)] model.language_model.layers.8.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
988
-> [양자화 X (원본유지)] model.language_model.layers.8.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
989
-> [양자화 X (원본유지)] model.language_model.layers.8.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
990
-> [양자화 O] model.language_model.layers.8.self_attn.k_proj.weight : 원본 형태 (512, 2048)
991
-> [양자화 O] model.language_model.layers.8.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
992
-> [양자화 X (원본유지)] model.language_model.layers.8.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
993
-> [양자화 O] model.language_model.layers.8.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
994
-> [양자화 O] model.language_model.layers.8.self_attn.v_proj.weight : 원본 형태 (512, 2048)
995
-> [양자화 X (원본유지)] model.language_model.layers.9.altup.correct_output_scale : 원본 형태 torch.Size([2048])
996
-> [양자화 X (원본유지)] model.language_model.layers.9.altup.correction_coefs.weight : 원본 형태 torch.Size([4, 4])
997
-> [양자화 X (원본유지)] model.language_model.layers.9.altup.modality_router.weight : 원본 형태 torch.Size([4, 2048])
998
-> [양자화 X (원본유지)] model.language_model.layers.9.altup.prediction_coefs.weight : 원본 형태 torch.Size([16, 4])
999
-> [양자화 X (원본유지)] model.language_model.layers.9.altup.router_norm.weight : 원본 형태 torch.Size([2048])
1000
-> [양자화 X (원본유지)] model.language_model.layers.9.input_layernorm.weight : 원본 형태 torch.Size([2048])
1001
-> [양자화 X (원본유지)] model.language_model.layers.9.laurel.linear_left.weight : 원본 형태 torch.Size([64, 2048])
1002
-> [양자화 X (원본유지)] model.language_model.layers.9.laurel.linear_right.weight : 원본 형태 torch.Size([2048, 64])
1003
-> [양자화 X (원본유지)] model.language_model.layers.9.laurel.post_laurel_norm.weight : 원본 형태 torch.Size([2048])
1004
-> [양자화 O] model.language_model.layers.9.mlp.down_proj.weight : 원본 형태 (2048, 8192)
1005
-> [양자화 O] model.language_model.layers.9.mlp.gate_proj.weight : 원본 형태 (8192, 2048)
1006
-> [양자화 O] model.language_model.layers.9.mlp.up_proj.weight : 원본 형태 (8192, 2048)
1007
-> [양자화 X (원본유지)] model.language_model.layers.9.per_layer_input_gate.weight : 원본 형태 torch.Size([256, 2048])
1008
-> [양자화 X (원본유지)] model.language_model.layers.9.per_layer_projection.weight : 원본 형태 torch.Size([2048, 256])
1009
-> [양자화 X (원본유지)] model.language_model.layers.9.post_attention_layernorm.weight : 원본 형태 torch.Size([2048])
1010
-> [양자화 X (원본유지)] model.language_model.layers.9.post_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
1011
-> [양자화 X (원본유지)] model.language_model.layers.9.post_per_layer_input_norm.weight : 원본 형태 torch.Size([2048])
1012
-> [양자화 X (원본유지)] model.language_model.layers.9.pre_feedforward_layernorm.weight : 원본 형태 torch.Size([2048])
1013
-> [양자화 X (원본유지)] model.language_model.layers.9.self_attn.k_norm.weight : 원본 형태 torch.Size([256])
1014
-> [양자화 O] model.language_model.layers.9.self_attn.k_proj.weight : 원본 형태 (512, 2048)
1015
-> [양자화 O] model.language_model.layers.9.self_attn.o_proj.weight : 원본 형태 (2048, 2048)
1016
-> [양자화 X (원본유지)] model.language_model.layers.9.self_attn.q_norm.weight : 원본 형태 torch.Size([256])
1017
-> [양자화 O] model.language_model.layers.9.self_attn.q_proj.weight : 원본 형태 (2048, 2048)
1018
-> [양자화 O] model.language_model.layers.9.self_attn.v_proj.weight : 원본 형태 (512, 2048)
1019
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.0.bn1.weight : 원본 형태 torch.Size([256])
1020
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.0.bn2.weight : 원본 형태 torch.Size([128])
1021
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.0.conv_exp.weight : 원본 형태 torch.Size([256, 64, 3, 3])
1022
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.0.conv_pwl.weight : 원본 형태 torch.Size([128, 256, 1, 1])
1023
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.1.bn1.weight : 원본 형태 torch.Size([512])
1024
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.1.bn2.weight : 원본 형태 torch.Size([128])
1025
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.1.conv_exp.weight : 원본 형태 torch.Size([512, 128, 3, 3])
1026
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.1.conv_pwl.weight : 원본 형태 torch.Size([128, 512, 1, 1])
1027
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.2.bn1.weight : 원본 형태 torch.Size([512])
1028
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.2.bn2.weight : 원본 형태 torch.Size([128])
1029
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.2.conv_exp.weight : 원본 형태 torch.Size([512, 128, 3, 3])
1030
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.0.2.conv_pwl.weight : 원본 형태 torch.Size([128, 512, 1, 1])
1031
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.dw_mid.bn.weight : 원본 형태 torch.Size([768])
1032
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.dw_mid.conv.weight : 원본 형태 torch.Size([768, 1, 5, 5])
1033
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.dw_start.bn.weight : 원본 형태 torch.Size([128])
1034
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.dw_start.conv.weight : 원본 형태 torch.Size([128, 1, 3, 3])
1035
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.layer_scale.gamma : 원본 형태 torch.Size([256])
1036
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.pw_exp.bn.weight : 원본 형태 torch.Size([768])
1037
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.pw_exp.conv.weight : 원본 형태 torch.Size([768, 128, 1, 1])
1038
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.pw_proj.bn.weight : 원본 형태 torch.Size([256])
1039
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.0.pw_proj.conv.weight : 원본 형태 torch.Size([256, 768, 1, 1])
1040
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.dw_start.bn.weight : 원본 형태 torch.Size([256])
1041
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.dw_start.conv.weight : 원본 형태 torch.Size([256, 1, 5, 5])
1042
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.layer_scale.gamma : 원본 형태 torch.Size([256])
1043
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.pw_exp.bn.weight : 원본 형태 torch.Size([1024])
1044
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.pw_exp.conv.weight : 원본 형태 torch.Size([1024, 256, 1, 1])
1045
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.pw_proj.bn.weight : 원본 형태 torch.Size([256])
1046
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.1.pw_proj.conv.weight : 원본 형태 torch.Size([256, 1024, 1, 1])
1047
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.dw_start.bn.weight : 원본 형태 torch.Size([256])
1048
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.dw_start.conv.weight : 원본 형태 torch.Size([256, 1, 3, 3])
1049
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.layer_scale.gamma : 원본 형태 torch.Size([256])
1050
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.pw_exp.bn.weight : 원본 형태 torch.Size([1024])
1051
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.pw_exp.conv.weight : 원본 형태 torch.Size([1024, 256, 1, 1])
1052
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.pw_proj.bn.weight : 원본 형태 torch.Size([256])
1053
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.2.pw_proj.conv.weight : 원본 형태 torch.Size([256, 1024, 1, 1])
1054
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.dw_start.bn.weight : 원본 형태 torch.Size([256])
1055
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.dw_start.conv.weight : 원본 형태 torch.Size([256, 1, 5, 5])
1056
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.layer_scale.gamma : 원본 형태 torch.Size([256])
1057
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.pw_exp.bn.weight : 원본 형태 torch.Size([1024])
1058
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.pw_exp.conv.weight : 원본 형태 torch.Size([1024, 256, 1, 1])
1059
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.pw_proj.bn.weight : 원본 형태 torch.Size([256])
1060
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.3.pw_proj.conv.weight : 원본 형태 torch.Size([256, 1024, 1, 1])
1061
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.dw_start.bn.weight : 원본 형태 torch.Size([256])
1062
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.dw_start.conv.weight : 원본 형태 torch.Size([256, 1, 3, 3])
1063
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.layer_scale.gamma : 원본 형태 torch.Size([256])
1064
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.pw_exp.bn.weight : 원본 형태 torch.Size([1024])
1065
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.pw_exp.conv.weight : 원본 형태 torch.Size([1024, 256, 1, 1])
1066
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.pw_proj.bn.weight : 원본 형태 torch.Size([256])
1067
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.1.4.pw_proj.conv.weight : 원본 형태 torch.Size([256, 1024, 1, 1])
1068
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.dw_mid.bn.weight : 원본 형태 torch.Size([1536])
1069
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.dw_mid.conv.weight : 원본 형태 torch.Size([1536, 1, 5, 5])
1070
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.dw_start.bn.weight : 원본 형태 torch.Size([256])
1071
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.dw_start.conv.weight : 원본 형태 torch.Size([256, 1, 5, 5])
1072
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.layer_scale.gamma : 원본 형태 torch.Size([640])
1073
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.pw_exp.bn.weight : 원본 형태 torch.Size([1536])
1074
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.pw_exp.conv.weight : 원본 형태 torch.Size([1536, 256, 1, 1])
1075
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1076
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.0.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1536, 1, 1])
1077
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.dw_start.bn.weight : 원본 형태 torch.Size([640])
1078
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1079
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.layer_scale.gamma : 원본 형태 torch.Size([640])
1080
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1081
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1082
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1083
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.1.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1084
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.10.layer_scale.gamma : 원본 형태 torch.Size([640])
1085
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.10.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1086
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.10.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1087
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.10.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1088
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.10.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1089
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1090
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.key.norm.weight : 원본 형태 torch.Size([640])
1091
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1092
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1093
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1094
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1095
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.value.norm.weight : 원본 형태 torch.Size([640])
1096
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1097
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.layer_scale.gamma : 원본 형태 torch.Size([640])
1098
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.11.norm.weight : 원본 형태 torch.Size([640])
1099
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.12.layer_scale.gamma : 원본 형태 torch.Size([640])
1100
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.12.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1101
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.12.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1102
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.12.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1103
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.12.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1104
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1105
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.key.norm.weight : 원본 형태 torch.Size([640])
1106
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1107
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1108
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1109
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1110
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.value.norm.weight : 원본 형태 torch.Size([640])
1111
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1112
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.layer_scale.gamma : 원본 형태 torch.Size([640])
1113
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.13.norm.weight : 원본 형태 torch.Size([640])
1114
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.14.layer_scale.gamma : 원본 형태 torch.Size([640])
1115
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.14.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1116
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.14.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1117
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.14.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1118
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.14.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1119
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1120
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.key.norm.weight : 원본 형태 torch.Size([640])
1121
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1122
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1123
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1124
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1125
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.value.norm.weight : 원본 형태 torch.Size([640])
1126
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1127
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.layer_scale.gamma : 원본 형태 torch.Size([640])
1128
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.15.norm.weight : 원본 형태 torch.Size([640])
1129
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.16.layer_scale.gamma : 원본 형태 torch.Size([640])
1130
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.16.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1131
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.16.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1132
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.16.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1133
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.16.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1134
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1135
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.key.norm.weight : 원본 형태 torch.Size([640])
1136
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1137
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1138
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1139
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1140
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.value.norm.weight : 원본 형태 torch.Size([640])
1141
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1142
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.layer_scale.gamma : 원본 형태 torch.Size([640])
1143
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.17.norm.weight : 원본 형태 torch.Size([640])
1144
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.18.layer_scale.gamma : 원본 형태 torch.Size([640])
1145
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.18.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1146
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.18.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1147
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.18.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1148
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.18.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1149
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1150
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.key.norm.weight : 원본 형태 torch.Size([640])
1151
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1152
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1153
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1154
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1155
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.value.norm.weight : 원본 형태 torch.Size([640])
1156
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1157
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.layer_scale.gamma : 원본 형태 torch.Size([640])
1158
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.19.norm.weight : 원본 형태 torch.Size([640])
1159
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.dw_start.bn.weight : 원본 형태 torch.Size([640])
1160
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1161
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.layer_scale.gamma : 원본 형태 torch.Size([640])
1162
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1163
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1164
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1165
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.2.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1166
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.20.layer_scale.gamma : 원본 형태 torch.Size([640])
1167
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.20.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1168
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.20.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1169
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.20.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1170
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.20.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1171
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1172
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.key.norm.weight : 원본 형태 torch.Size([640])
1173
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1174
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1175
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1176
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1177
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.value.norm.weight : 원본 형태 torch.Size([640])
1178
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1179
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.layer_scale.gamma : 원본 형태 torch.Size([640])
1180
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.21.norm.weight : 원본 형태 torch.Size([640])
1181
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.22.layer_scale.gamma : 원본 형태 torch.Size([640])
1182
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.22.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1183
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.22.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1184
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.22.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1185
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.22.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1186
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1187
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.key.norm.weight : 원본 형태 torch.Size([640])
1188
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1189
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1190
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1191
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1192
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.value.norm.weight : 원본 형태 torch.Size([640])
1193
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1194
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.layer_scale.gamma : 원본 형태 torch.Size([640])
1195
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.23.norm.weight : 원본 형태 torch.Size([640])
1196
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.24.layer_scale.gamma : 원본 형태 torch.Size([640])
1197
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.24.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1198
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.24.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1199
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.24.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1200
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.24.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1201
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1202
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.key.norm.weight : 원본 형태 torch.Size([640])
1203
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1204
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1205
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1206
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1207
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.value.norm.weight : 원본 형태 torch.Size([640])
1208
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1209
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.layer_scale.gamma : 원본 형태 torch.Size([640])
1210
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.25.norm.weight : 원본 형태 torch.Size([640])
1211
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.26.layer_scale.gamma : 원본 형태 torch.Size([640])
1212
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.26.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1213
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.26.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1214
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.26.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1215
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.26.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1216
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1217
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.key.norm.weight : 원본 형태 torch.Size([640])
1218
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1219
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1220
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1221
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1222
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.value.norm.weight : 원본 형태 torch.Size([640])
1223
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1224
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.layer_scale.gamma : 원본 형태 torch.Size([640])
1225
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.27.norm.weight : 원본 형태 torch.Size([640])
1226
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.28.layer_scale.gamma : 원본 형태 torch.Size([640])
1227
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.28.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1228
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.28.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1229
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.28.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1230
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.28.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1231
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1232
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.key.norm.weight : 원본 형태 torch.Size([640])
1233
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1234
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1235
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1236
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1237
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.value.norm.weight : 원본 형태 torch.Size([640])
1238
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1239
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.layer_scale.gamma : 원본 형태 torch.Size([640])
1240
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.29.norm.weight : 원본 형태 torch.Size([640])
1241
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.dw_start.bn.weight : 원본 형태 torch.Size([640])
1242
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1243
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.layer_scale.gamma : 원본 형태 torch.Size([640])
1244
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1245
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1246
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1247
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.3.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1248
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.30.layer_scale.gamma : 원본 형태 torch.Size([640])
1249
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.30.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1250
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.30.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1251
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.30.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1252
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.30.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1253
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1254
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.key.norm.weight : 원본 형태 torch.Size([640])
1255
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1256
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1257
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1258
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1259
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.value.norm.weight : 원본 형태 torch.Size([640])
1260
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1261
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.layer_scale.gamma : 원본 형태 torch.Size([640])
1262
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.31.norm.weight : 원본 형태 torch.Size([640])
1263
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.32.layer_scale.gamma : 원본 형태 torch.Size([640])
1264
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.32.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1265
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.32.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1266
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.32.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1267
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.32.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1268
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1269
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.key.norm.weight : 원본 형태 torch.Size([640])
1270
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1271
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1272
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1273
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1274
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.value.norm.weight : 원본 형태 torch.Size([640])
1275
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1276
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.layer_scale.gamma : 원본 형태 torch.Size([640])
1277
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.33.norm.weight : 원본 형태 torch.Size([640])
1278
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.34.layer_scale.gamma : 원본 형태 torch.Size([640])
1279
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.34.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1280
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.34.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1281
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.34.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1282
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.34.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1283
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1284
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.key.norm.weight : 원본 형태 torch.Size([640])
1285
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1286
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1287
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1288
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1289
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.value.norm.weight : 원본 형태 torch.Size([640])
1290
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1291
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.layer_scale.gamma : 원본 형태 torch.Size([640])
1292
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.35.norm.weight : 원본 형태 torch.Size([640])
1293
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.36.layer_scale.gamma : 원본 형태 torch.Size([640])
1294
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.36.pw_exp.bn.weight : 원본 형태 torch.Size([1280])
1295
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.36.pw_exp.conv.weight : 원본 형태 torch.Size([1280, 640, 1, 1])
1296
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.36.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1297
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.36.pw_proj.conv.weight : 원본 형태 torch.Size([640, 1280, 1, 1])
1298
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.dw_start.bn.weight : 원본 형태 torch.Size([640])
1299
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1300
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.layer_scale.gamma : 원본 형태 torch.Size([640])
1301
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1302
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1303
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1304
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.4.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1305
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.dw_start.bn.weight : 원본 형태 torch.Size([640])
1306
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1307
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.layer_scale.gamma : 원본 형태 torch.Size([640])
1308
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1309
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1310
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1311
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.5.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1312
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.dw_start.bn.weight : 원본 형태 torch.Size([640])
1313
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1314
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.layer_scale.gamma : 원본 형태 torch.Size([640])
1315
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1316
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1317
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1318
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.6.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1319
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.dw_start.bn.weight : 원본 형태 torch.Size([640])
1320
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1321
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.layer_scale.gamma : 원본 형태 torch.Size([640])
1322
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1323
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 640, 1, 1])
1324
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1325
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.7.pw_proj.conv.weight : 원본 형태 torch.Size([640, 2560, 1, 1])
1326
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.8.layer_scale.gamma : 원본 형태 torch.Size([640])
1327
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.8.pw_exp.bn.weight : 원본 형태 torch.Size([640])
1328
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.8.pw_exp.conv.weight : 원본 형태 torch.Size([640, 640, 1, 1])
1329
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.8.pw_proj.bn.weight : 원본 형태 torch.Size([640])
1330
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.8.pw_proj.conv.weight : 원본 형태 torch.Size([640, 640, 1, 1])
1331
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.key.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1332
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.key.norm.weight : 원본 형태 torch.Size([640])
1333
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.key.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1334
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.output.proj.weight : 원본 형태 torch.Size([640, 768, 1, 1])
1335
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.query.proj.weight : 원본 형태 torch.Size([768, 640, 1, 1])
1336
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.value.down_conv.weight : 원본 형태 torch.Size([640, 1, 3, 3])
1337
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.value.norm.weight : 원본 형태 torch.Size([640])
1338
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.attn.value.proj.weight : 원본 형태 torch.Size([64, 640, 1, 1])
1339
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.layer_scale.gamma : 원본 형태 torch.Size([640])
1340
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.2.9.norm.weight : 원본 형태 torch.Size([640])
1341
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.dw_mid.bn.weight : 원본 형태 torch.Size([3840])
1342
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.dw_mid.conv.weight : 원본 형태 torch.Size([3840, 1, 5, 5])
1343
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.dw_start.bn.weight : 원본 형태 torch.Size([640])
1344
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.dw_start.conv.weight : 원본 형태 torch.Size([640, 1, 5, 5])
1345
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.layer_scale.gamma : 원본 형태 torch.Size([1280])
1346
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.pw_exp.bn.weight : 원본 형태 torch.Size([3840])
1347
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.pw_exp.conv.weight : 원본 형태 torch.Size([3840, 640, 1, 1])
1348
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1349
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.0.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 3840, 1, 1])
1350
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1351
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1352
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1353
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1354
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.layer_scale.gamma : 원본 형태 torch.Size([1280])
1355
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.1.norm.weight : 원본 형태 torch.Size([1280])
1356
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.10.layer_scale.gamma : 원본 형태 torch.Size([1280])
1357
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.10.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1358
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.10.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1359
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.10.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1360
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.10.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1361
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1362
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1363
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1364
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1365
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.layer_scale.gamma : 원본 형태 torch.Size([1280])
1366
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.11.norm.weight : 원본 형태 torch.Size([1280])
1367
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.12.layer_scale.gamma : 원본 형태 torch.Size([1280])
1368
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.12.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1369
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.12.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1370
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.12.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1371
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.12.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1372
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1373
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1374
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1375
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1376
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.layer_scale.gamma : 원본 형태 torch.Size([1280])
1377
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.13.norm.weight : 원본 형태 torch.Size([1280])
1378
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.14.layer_scale.gamma : 원본 형태 torch.Size([1280])
1379
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.14.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1380
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.14.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1381
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.14.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1382
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.14.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1383
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1384
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1385
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1386
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1387
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.layer_scale.gamma : 원본 형태 torch.Size([1280])
1388
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.15.norm.weight : 원본 형태 torch.Size([1280])
1389
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.16.layer_scale.gamma : 원본 형태 torch.Size([1280])
1390
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.16.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1391
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.16.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1392
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.16.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1393
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.16.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1394
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1395
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1396
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1397
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1398
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.layer_scale.gamma : 원본 형태 torch.Size([1280])
1399
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.17.norm.weight : 원본 형태 torch.Size([1280])
1400
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.18.layer_scale.gamma : 원본 형태 torch.Size([1280])
1401
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.18.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1402
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.18.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1403
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.18.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1404
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.18.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1405
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1406
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1407
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1408
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1409
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.layer_scale.gamma : 원본 형태 torch.Size([1280])
1410
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.19.norm.weight : 원본 형태 torch.Size([1280])
1411
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.2.layer_scale.gamma : 원본 형태 torch.Size([1280])
1412
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.2.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1413
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.2.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1414
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.2.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1415
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.2.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1416
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.20.layer_scale.gamma : 원본 형태 torch.Size([1280])
1417
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.20.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1418
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.20.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1419
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.20.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1420
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.20.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1421
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1422
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1423
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1424
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1425
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.layer_scale.gamma : 원본 형태 torch.Size([1280])
1426
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.21.norm.weight : 원본 형태 torch.Size([1280])
1427
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.22.layer_scale.gamma : 원본 형태 torch.Size([1280])
1428
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.22.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1429
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.22.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1430
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.22.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1431
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.22.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1432
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1433
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1434
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1435
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1436
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.layer_scale.gamma : 원본 형태 torch.Size([1280])
1437
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.23.norm.weight : 원본 형태 torch.Size([1280])
1438
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.24.layer_scale.gamma : 원본 형태 torch.Size([1280])
1439
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.24.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1440
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.24.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1441
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.24.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1442
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.24.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1443
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1444
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1445
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1446
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1447
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.layer_scale.gamma : 원본 형태 torch.Size([1280])
1448
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.25.norm.weight : 원본 형태 torch.Size([1280])
1449
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.26.layer_scale.gamma : 원본 형태 torch.Size([1280])
1450
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.26.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1451
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.26.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1452
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.26.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1453
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.26.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1454
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1455
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1456
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1457
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1458
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.layer_scale.gamma : 원본 형태 torch.Size([1280])
1459
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.27.norm.weight : 원본 형태 torch.Size([1280])
1460
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.28.layer_scale.gamma : 원본 형태 torch.Size([1280])
1461
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.28.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1462
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.28.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1463
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.28.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1464
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.28.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1465
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1466
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1467
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1468
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1469
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.layer_scale.gamma : 원본 형태 torch.Size([1280])
1470
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.29.norm.weight : 원본 형태 torch.Size([1280])
1471
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1472
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1473
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1474
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1475
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.layer_scale.gamma : 원본 형태 torch.Size([1280])
1476
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.3.norm.weight : 원본 형태 torch.Size([1280])
1477
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.30.layer_scale.gamma : 원본 형태 torch.Size([1280])
1478
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.30.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1479
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.30.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1480
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.30.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1481
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.30.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1482
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1483
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1484
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1485
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1486
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.layer_scale.gamma : 원본 형태 torch.Size([1280])
1487
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.31.norm.weight : 원본 형태 torch.Size([1280])
1488
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.32.layer_scale.gamma : 원본 형태 torch.Size([1280])
1489
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.32.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1490
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.32.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1491
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.32.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1492
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.32.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1493
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1494
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1495
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1496
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1497
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.layer_scale.gamma : 원본 형태 torch.Size([1280])
1498
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.33.norm.weight : 원본 형태 torch.Size([1280])
1499
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.34.layer_scale.gamma : 원본 형태 torch.Size([1280])
1500
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.34.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1501
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.34.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1502
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.34.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1503
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.34.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1504
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1505
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1506
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1507
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1508
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.layer_scale.gamma : 원본 형태 torch.Size([1280])
1509
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.35.norm.weight : 원본 형태 torch.Size([1280])
1510
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.36.layer_scale.gamma : 원본 형태 torch.Size([1280])
1511
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.36.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1512
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.36.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1513
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.36.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1514
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.36.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1515
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1516
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1517
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1518
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1519
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.layer_scale.gamma : 원본 형태 torch.Size([1280])
1520
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.37.norm.weight : 원본 형태 torch.Size([1280])
1521
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.38.layer_scale.gamma : 원본 형태 torch.Size([1280])
1522
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.38.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1523
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.38.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1524
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.38.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1525
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.38.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1526
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.4.layer_scale.gamma : 원본 형태 torch.Size([1280])
1527
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.4.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1528
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.4.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1529
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.4.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1530
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.4.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1531
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1532
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1533
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1534
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1535
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.layer_scale.gamma : 원본 형태 torch.Size([1280])
1536
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.5.norm.weight : 원본 형태 torch.Size([1280])
1537
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.6.layer_scale.gamma : 원본 형태 torch.Size([1280])
1538
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.6.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1539
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.6.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1540
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.6.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1541
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.6.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1542
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1543
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1544
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1545
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1546
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.layer_scale.gamma : 원본 형태 torch.Size([1280])
1547
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.7.norm.weight : 원본 형태 torch.Size([1280])
1548
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.8.layer_scale.gamma : 원본 형태 torch.Size([1280])
1549
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.8.pw_exp.bn.weight : 원본 형태 torch.Size([2560])
1550
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.8.pw_exp.conv.weight : 원본 형태 torch.Size([2560, 1280, 1, 1])
1551
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.8.pw_proj.bn.weight : 원본 형태 torch.Size([1280])
1552
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.8.pw_proj.conv.weight : 원본 형태 torch.Size([1280, 2560, 1, 1])
1553
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.attn.key.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1554
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.attn.output.proj.weight : 원본 형태 torch.Size([1280, 1536, 1, 1])
1555
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.attn.query.proj.weight : 원본 형태 torch.Size([1536, 1280, 1, 1])
1556
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.attn.value.proj.weight : 원본 형태 torch.Size([96, 1280, 1, 1])
1557
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.layer_scale.gamma : 원본 형태 torch.Size([1280])
1558
-> [양자화 X (원본유지)] model.vision_tower.timm_model.blocks.3.9.norm.weight : 원본 형태 torch.Size([1280])
1559
-> [양자화 X (원본유지)] model.vision_tower.timm_model.conv_stem.bn.weight : 원본 형태 torch.Size([64])
1560
-> [양자화 X (원본유지)] model.vision_tower.timm_model.conv_stem.conv.weight : 원본 형태 torch.Size([64, 3, 3, 3])
1561
-> [양자화 X (원본유지)] model.vision_tower.timm_model.msfa.ffn.pw_exp.bn.weight : 원본 형태 torch.Size([3840])
1562
-> [양자화 X (원본유지)] model.vision_tower.timm_model.msfa.ffn.pw_exp.conv.weight : 원본 형태 torch.Size([3840, 1920, 1, 1])
1563
-> [양자화 X (원본유지)] model.vision_tower.timm_model.msfa.ffn.pw_proj.bn.weight : 원본 형태 torch.Size([2048])
1564
-> [양자화 X (원본유지)] model.vision_tower.timm_model.msfa.ffn.pw_proj.conv.weight : 원본 형태 torch.Size([2048, 3840, 1, 1])
1565
-> [양자화 X (원본유지)] model.vision_tower.timm_model.msfa.norm.weight : 원본 형태 torch.Size([2048])