Hugging Face's logo

Disty0
/
LTX-2-SDNQ-4bit-dynamic

Diffusers
Safetensors
LTX2Pipeline
sdnq
ltx-2
4-bit precision
Model card Files
xet
Community
3
Disty0 commited on
Commit
c366ccb
·
verified ·
1 Parent(s): 1a526e1

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +1 -0
  2. audio_vae/config.json +26 -0
  3. audio_vae/diffusion_pytorch_model.safetensors +3 -0
  4. connectors/config.json +20 -0
  5. connectors/diffusion_pytorch_model.safetensors +3 -0
  6. latent_upsampler/config.json +11 -0
  7. latent_upsampler/diffusion_pytorch_model.safetensors +3 -0
  8. model_index.json +37 -0
  9. scheduler/scheduler_config.json +18 -0
  10. text_encoder/config.json +672 -0
  11. text_encoder/generation_config.json +11 -0
  12. text_encoder/model-00001-of-00002.safetensors +3 -0
  13. text_encoder/model-00002-of-00002.safetensors +3 -0
  14. text_encoder/model.safetensors.index.json +0 -0
  15. text_encoder/quantization_config.json +558 -0
  16. tokenizer/added_tokens.json +3 -0
  17. tokenizer/chat_template.jinja +47 -0
  18. tokenizer/special_tokens_map.json +33 -0
  19. tokenizer/tokenizer.json +3 -0
  20. tokenizer/tokenizer.model +3 -0
  21. tokenizer/tokenizer_config.json +0 -0
  22. transformer/config.json +1436 -0
  23. transformer/diffusion_pytorch_model-00001-of-00003.safetensors +3 -0
  24. transformer/diffusion_pytorch_model-00002-of-00003.safetensors +3 -0
  25. transformer/diffusion_pytorch_model-00003-of-00003.safetensors +3 -0
  26. transformer/diffusion_pytorch_model.safetensors.index.json +0 -0
  27. transformer/quantization_config.json +1391 -0
  28. vae/config.json +82 -0
  29. vae/diffusion_pytorch_model.safetensors +3 -0
  30. vocoder/config.json +46 -0
  31. vocoder/diffusion_pytorch_model.safetensors +3 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
audio_vae/config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "AutoencoderKLLTX2Audio",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "/home/disty/.cache/huggingface/hub/models--Lightricks--LTX-2/snapshots/64f3f196f34e3deb03c85333ec34e4c9d9f82593/audio_vae",
5
+ "attn_resolutions": null,
6
+ "base_channels": 128,
7
+ "causality_axis": "height",
8
+ "ch_mult": [
9
+ 1,
10
+ 2,
11
+ 4
12
+ ],
13
+ "double_z": true,
14
+ "dropout": 0.0,
15
+ "in_channels": 2,
16
+ "is_causal": true,
17
+ "latent_channels": 8,
18
+ "mel_bins": 64,
19
+ "mel_hop_length": 160,
20
+ "mid_block_add_attention": false,
21
+ "norm_type": "pixel",
22
+ "num_res_blocks": 2,
23
+ "output_channels": 2,
24
+ "resolution": 256,
25
+ "sample_rate": 16000
26
+ }
audio_vae/diffusion_pytorch_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b36ce4066065ce0aa5ff4d7cf96a3e3bc0859c6fefcf864663e3fe686c5c181c
3
+ size 106507972
connectors/config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "LTX2TextConnectors",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "/home/disty/.cache/huggingface/hub/models--Lightricks--LTX-2/snapshots/64f3f196f34e3deb03c85333ec34e4c9d9f82593/connectors",
5
+ "audio_connector_attention_head_dim": 128,
6
+ "audio_connector_num_attention_heads": 30,
7
+ "audio_connector_num_layers": 2,
8
+ "audio_connector_num_learnable_registers": 128,
9
+ "caption_channels": 3840,
10
+ "causal_temporal_positioning": false,
11
+ "connector_rope_base_seq_len": 4096,
12
+ "rope_double_precision": true,
13
+ "rope_theta": 10000.0,
14
+ "rope_type": "split",
15
+ "text_proj_in_factor": 49,
16
+ "video_connector_attention_head_dim": 128,
17
+ "video_connector_num_attention_heads": 30,
18
+ "video_connector_num_layers": 2,
19
+ "video_connector_num_learnable_registers": 128
20
+ }
connectors/diffusion_pytorch_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7c0ad36c2d0706fb229193d5c698f0ef50c9b33678140b4ee84723a047b4032
3
+ size 2862957976
latent_upsampler/config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "LTX2LatentUpsamplerModel",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "dims": 3,
5
+ "in_channels": 128,
6
+ "mid_channels": 1024,
7
+ "num_blocks_per_stage": 4,
8
+ "rational_spatial_scale": 2.0,
9
+ "spatial_upsample": true,
10
+ "temporal_upsample": false
11
+ }
latent_upsampler/diffusion_pytorch_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c56276acbffb30f97824b4c2fd6770e8157d65e5be7a93e2307393c1ebbb1f12
3
+ size 995743482
model_index.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "LTX2Pipeline",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "Lightricks/LTX-2",
5
+ "audio_vae": [
6
+ "diffusers",
7
+ "AutoencoderKLLTX2Audio"
8
+ ],
9
+ "connectors": [
10
+ "ltx2",
11
+ "LTX2TextConnectors"
12
+ ],
13
+ "scheduler": [
14
+ "diffusers",
15
+ "FlowMatchEulerDiscreteScheduler"
16
+ ],
17
+ "text_encoder": [
18
+ "transformers",
19
+ "Gemma3ForConditionalGeneration"
20
+ ],
21
+ "tokenizer": [
22
+ "transformers",
23
+ "GemmaTokenizerFast"
24
+ ],
25
+ "transformer": [
26
+ "diffusers",
27
+ "LTX2VideoTransformer3DModel"
28
+ ],
29
+ "vae": [
30
+ "diffusers",
31
+ "AutoencoderKLLTX2Video"
32
+ ],
33
+ "vocoder": [
34
+ "ltx2",
35
+ "LTX2Vocoder"
36
+ ]
37
+ }
scheduler/scheduler_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "FlowMatchEulerDiscreteScheduler",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "base_image_seq_len": 1024,
5
+ "base_shift": 0.95,
6
+ "invert_sigmas": false,
7
+ "max_image_seq_len": 4096,
8
+ "max_shift": 2.05,
9
+ "num_train_timesteps": 1000,
10
+ "shift": 1.0,
11
+ "shift_terminal": 0.1,
12
+ "stochastic_sampling": false,
13
+ "time_shift_type": "exponential",
14
+ "use_beta_sigmas": false,
15
+ "use_dynamic_shifting": true,
16
+ "use_exponential_sigmas": false,
17
+ "use_karras_sigmas": false
18
+ }
text_encoder/config.json ADDED
@@ -0,0 +1,672 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
+ ],
5
+ "boi_token_index": 255999,
6
+ "dtype": "bfloat16",
7
+ "eoi_token_index": 256000,
8
+ "eos_token_id": [
9
+ 1,
10
+ 106
11
+ ],
12
+ "image_token_index": 262144,
13
+ "initializer_range": 0.02,
14
+ "mm_tokens_per_image": 256,
15
+ "model_type": "gemma3",
16
+ "quantization_config": {
17
+ "add_skip_keys": false,
18
+ "dequantize_fp32": false,
19
+ "dynamic_loss_threshold": 0.01,
20
+ "group_size": 0,
21
+ "is_integer": true,
22
+ "is_training": false,
23
+ "modules_dtype_dict": {
24
+ "int5": [
25
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight",
26
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight",
27
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight",
28
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight",
29
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight",
30
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight",
31
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight",
32
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight",
33
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight",
34
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight",
35
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight",
36
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight",
37
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight",
38
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight",
39
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight",
40
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight",
41
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight",
42
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight",
43
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight",
44
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight",
45
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight",
46
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight",
47
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight",
48
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight",
49
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight",
50
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight",
51
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight",
52
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight",
53
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight",
54
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight",
55
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight",
56
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight",
57
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight",
58
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight",
59
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight",
60
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight",
61
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight",
62
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight",
63
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight",
64
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight",
65
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight",
66
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight",
67
+ "model.language_model.layers.17.self_attn.k_proj.weight"
68
+ ],
69
+ "uint4": [
70
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight",
71
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight",
72
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight",
73
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight",
74
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight",
75
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight",
76
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight",
77
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight",
78
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight",
79
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight",
80
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight",
81
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight",
82
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight",
83
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight",
84
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight",
85
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight",
86
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight",
87
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight",
88
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight",
89
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight",
90
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight",
91
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight",
92
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight",
93
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight",
94
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight",
95
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight",
96
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight",
97
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight",
98
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight",
99
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight",
100
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight",
101
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight",
102
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight",
103
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight",
104
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight",
105
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight",
106
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight",
107
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight",
108
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight",
109
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight",
110
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight",
111
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight",
112
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight",
113
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight",
114
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight",
115
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight",
116
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight",
117
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight",
118
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight",
119
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight",
120
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight",
121
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight",
122
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight",
123
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight",
124
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight",
125
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight",
126
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight",
127
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight",
128
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight",
129
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight",
130
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight",
131
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight",
132
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight",
133
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight",
134
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight",
135
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight",
136
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight",
137
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight",
138
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight",
139
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight",
140
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight",
141
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight",
142
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight",
143
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight",
144
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight",
145
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight",
146
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight",
147
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight",
148
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight",
149
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight",
150
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight",
151
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight",
152
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight",
153
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight",
154
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight",
155
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight",
156
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight",
157
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight",
158
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight",
159
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight",
160
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight",
161
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight",
162
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight",
163
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight",
164
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight",
165
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight",
166
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight",
167
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight",
168
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight",
169
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight",
170
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight",
171
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight",
172
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight",
173
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight",
174
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight",
175
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight",
176
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight",
177
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight",
178
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight",
179
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight",
180
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight",
181
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight",
182
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight",
183
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight",
184
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight",
185
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight",
186
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight",
187
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight",
188
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight",
189
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight",
190
+ "model.language_model.layers.0.mlp.down_proj.weight",
191
+ "model.language_model.layers.0.mlp.gate_proj.weight",
192
+ "model.language_model.layers.0.mlp.up_proj.weight",
193
+ "model.language_model.layers.0.self_attn.k_proj.weight",
194
+ "model.language_model.layers.0.self_attn.o_proj.weight",
195
+ "model.language_model.layers.0.self_attn.q_proj.weight",
196
+ "model.language_model.layers.0.self_attn.v_proj.weight",
197
+ "model.language_model.layers.1.self_attn.q_proj.weight",
198
+ "model.language_model.layers.1.mlp.down_proj.weight",
199
+ "model.language_model.layers.1.mlp.gate_proj.weight",
200
+ "model.language_model.layers.1.mlp.up_proj.weight",
201
+ "model.language_model.layers.1.self_attn.k_proj.weight",
202
+ "model.language_model.layers.1.self_attn.o_proj.weight",
203
+ "model.language_model.layers.1.self_attn.v_proj.weight",
204
+ "model.language_model.layers.2.mlp.down_proj.weight",
205
+ "model.language_model.layers.2.mlp.gate_proj.weight",
206
+ "model.language_model.layers.2.mlp.up_proj.weight",
207
+ "model.language_model.layers.2.self_attn.k_proj.weight",
208
+ "model.language_model.layers.2.self_attn.o_proj.weight",
209
+ "model.language_model.layers.2.self_attn.q_proj.weight",
210
+ "model.language_model.layers.2.self_attn.v_proj.weight",
211
+ "model.language_model.layers.3.mlp.down_proj.weight",
212
+ "model.language_model.layers.3.mlp.gate_proj.weight",
213
+ "model.language_model.layers.3.mlp.up_proj.weight",
214
+ "model.language_model.layers.3.self_attn.k_proj.weight",
215
+ "model.language_model.layers.3.self_attn.o_proj.weight",
216
+ "model.language_model.layers.3.self_attn.q_proj.weight",
217
+ "model.language_model.layers.3.self_attn.v_proj.weight",
218
+ "model.language_model.layers.4.mlp.down_proj.weight",
219
+ "model.language_model.layers.4.mlp.gate_proj.weight",
220
+ "model.language_model.layers.4.mlp.up_proj.weight",
221
+ "model.language_model.layers.4.self_attn.k_proj.weight",
222
+ "model.language_model.layers.4.self_attn.o_proj.weight",
223
+ "model.language_model.layers.4.self_attn.q_proj.weight",
224
+ "model.language_model.layers.4.self_attn.v_proj.weight",
225
+ "model.language_model.layers.5.mlp.down_proj.weight",
226
+ "model.language_model.layers.5.mlp.gate_proj.weight",
227
+ "model.language_model.layers.5.mlp.up_proj.weight",
228
+ "model.language_model.layers.5.self_attn.k_proj.weight",
229
+ "model.language_model.layers.5.self_attn.o_proj.weight",
230
+ "model.language_model.layers.5.self_attn.q_proj.weight",
231
+ "model.language_model.layers.5.self_attn.v_proj.weight",
232
+ "model.language_model.layers.6.mlp.gate_proj.weight",
233
+ "model.language_model.layers.6.self_attn.k_proj.weight",
234
+ "model.language_model.layers.6.self_attn.o_proj.weight",
235
+ "model.language_model.layers.6.self_attn.q_proj.weight",
236
+ "model.language_model.layers.6.self_attn.v_proj.weight",
237
+ "model.language_model.layers.10.mlp.down_proj.weight",
238
+ "model.language_model.layers.10.mlp.gate_proj.weight",
239
+ "model.language_model.layers.10.mlp.up_proj.weight",
240
+ "model.language_model.layers.10.self_attn.k_proj.weight",
241
+ "model.language_model.layers.10.self_attn.o_proj.weight",
242
+ "model.language_model.layers.10.self_attn.q_proj.weight",
243
+ "model.language_model.layers.10.self_attn.v_proj.weight",
244
+ "model.language_model.layers.11.mlp.down_proj.weight",
245
+ "model.language_model.layers.11.mlp.gate_proj.weight",
246
+ "model.language_model.layers.11.mlp.up_proj.weight",
247
+ "model.language_model.layers.11.self_attn.k_proj.weight",
248
+ "model.language_model.layers.11.self_attn.o_proj.weight",
249
+ "model.language_model.layers.11.self_attn.q_proj.weight",
250
+ "model.language_model.layers.11.self_attn.v_proj.weight",
251
+ "model.language_model.layers.6.mlp.down_proj.weight",
252
+ "model.language_model.layers.6.mlp.up_proj.weight",
253
+ "model.language_model.layers.7.mlp.down_proj.weight",
254
+ "model.language_model.layers.7.mlp.gate_proj.weight",
255
+ "model.language_model.layers.7.mlp.up_proj.weight",
256
+ "model.language_model.layers.7.self_attn.k_proj.weight",
257
+ "model.language_model.layers.7.self_attn.o_proj.weight",
258
+ "model.language_model.layers.7.self_attn.q_proj.weight",
259
+ "model.language_model.layers.7.self_attn.v_proj.weight",
260
+ "model.language_model.layers.8.mlp.down_proj.weight",
261
+ "model.language_model.layers.8.mlp.gate_proj.weight",
262
+ "model.language_model.layers.8.mlp.up_proj.weight",
263
+ "model.language_model.layers.8.self_attn.k_proj.weight",
264
+ "model.language_model.layers.8.self_attn.o_proj.weight",
265
+ "model.language_model.layers.8.self_attn.q_proj.weight",
266
+ "model.language_model.layers.8.self_attn.v_proj.weight",
267
+ "model.language_model.layers.9.mlp.down_proj.weight",
268
+ "model.language_model.layers.9.mlp.gate_proj.weight",
269
+ "model.language_model.layers.9.mlp.up_proj.weight",
270
+ "model.language_model.layers.9.self_attn.k_proj.weight",
271
+ "model.language_model.layers.9.self_attn.o_proj.weight",
272
+ "model.language_model.layers.9.self_attn.q_proj.weight",
273
+ "model.language_model.layers.9.self_attn.v_proj.weight",
274
+ "model.language_model.layers.12.mlp.down_proj.weight",
275
+ "model.language_model.layers.12.mlp.gate_proj.weight",
276
+ "model.language_model.layers.12.mlp.up_proj.weight",
277
+ "model.language_model.layers.12.self_attn.k_proj.weight",
278
+ "model.language_model.layers.12.self_attn.o_proj.weight",
279
+ "model.language_model.layers.12.self_attn.q_proj.weight",
280
+ "model.language_model.layers.12.self_attn.v_proj.weight",
281
+ "model.language_model.layers.13.mlp.down_proj.weight",
282
+ "model.language_model.layers.13.mlp.gate_proj.weight",
283
+ "model.language_model.layers.13.mlp.up_proj.weight",
284
+ "model.language_model.layers.13.self_attn.k_proj.weight",
285
+ "model.language_model.layers.13.self_attn.o_proj.weight",
286
+ "model.language_model.layers.13.self_attn.q_proj.weight",
287
+ "model.language_model.layers.13.self_attn.v_proj.weight",
288
+ "model.language_model.layers.14.mlp.down_proj.weight",
289
+ "model.language_model.layers.14.mlp.gate_proj.weight",
290
+ "model.language_model.layers.14.mlp.up_proj.weight",
291
+ "model.language_model.layers.14.self_attn.k_proj.weight",
292
+ "model.language_model.layers.14.self_attn.o_proj.weight",
293
+ "model.language_model.layers.14.self_attn.q_proj.weight",
294
+ "model.language_model.layers.14.self_attn.v_proj.weight",
295
+ "model.language_model.layers.15.mlp.down_proj.weight",
296
+ "model.language_model.layers.15.mlp.gate_proj.weight",
297
+ "model.language_model.layers.15.mlp.up_proj.weight",
298
+ "model.language_model.layers.15.self_attn.k_proj.weight",
299
+ "model.language_model.layers.15.self_attn.o_proj.weight",
300
+ "model.language_model.layers.15.self_attn.q_proj.weight",
301
+ "model.language_model.layers.15.self_attn.v_proj.weight",
302
+ "model.language_model.layers.16.mlp.down_proj.weight",
303
+ "model.language_model.layers.16.mlp.gate_proj.weight",
304
+ "model.language_model.layers.16.mlp.up_proj.weight",
305
+ "model.language_model.layers.16.self_attn.k_proj.weight",
306
+ "model.language_model.layers.16.self_attn.o_proj.weight",
307
+ "model.language_model.layers.16.self_attn.q_proj.weight",
308
+ "model.language_model.layers.16.self_attn.v_proj.weight",
309
+ "model.language_model.layers.17.mlp.gate_proj.weight",
310
+ "model.language_model.layers.17.self_attn.o_proj.weight",
311
+ "model.language_model.layers.17.self_attn.q_proj.weight",
312
+ "model.language_model.layers.17.self_attn.v_proj.weight",
313
+ "model.language_model.layers.17.mlp.down_proj.weight",
314
+ "model.language_model.layers.17.mlp.up_proj.weight",
315
+ "model.language_model.layers.18.mlp.down_proj.weight",
316
+ "model.language_model.layers.18.mlp.gate_proj.weight",
317
+ "model.language_model.layers.18.mlp.up_proj.weight",
318
+ "model.language_model.layers.18.self_attn.k_proj.weight",
319
+ "model.language_model.layers.18.self_attn.o_proj.weight",
320
+ "model.language_model.layers.18.self_attn.q_proj.weight",
321
+ "model.language_model.layers.18.self_attn.v_proj.weight",
322
+ "model.language_model.layers.19.mlp.down_proj.weight",
323
+ "model.language_model.layers.19.mlp.gate_proj.weight",
324
+ "model.language_model.layers.19.mlp.up_proj.weight",
325
+ "model.language_model.layers.19.self_attn.k_proj.weight",
326
+ "model.language_model.layers.19.self_attn.o_proj.weight",
327
+ "model.language_model.layers.19.self_attn.q_proj.weight",
328
+ "model.language_model.layers.19.self_attn.v_proj.weight",
329
+ "model.language_model.layers.20.mlp.down_proj.weight",
330
+ "model.language_model.layers.20.mlp.gate_proj.weight",
331
+ "model.language_model.layers.20.mlp.up_proj.weight",
332
+ "model.language_model.layers.20.self_attn.k_proj.weight",
333
+ "model.language_model.layers.20.self_attn.o_proj.weight",
334
+ "model.language_model.layers.20.self_attn.q_proj.weight",
335
+ "model.language_model.layers.20.self_attn.v_proj.weight",
336
+ "model.language_model.layers.21.mlp.down_proj.weight",
337
+ "model.language_model.layers.21.mlp.gate_proj.weight",
338
+ "model.language_model.layers.21.mlp.up_proj.weight",
339
+ "model.language_model.layers.21.self_attn.k_proj.weight",
340
+ "model.language_model.layers.21.self_attn.o_proj.weight",
341
+ "model.language_model.layers.21.self_attn.q_proj.weight",
342
+ "model.language_model.layers.21.self_attn.v_proj.weight",
343
+ "model.language_model.layers.22.mlp.down_proj.weight",
344
+ "model.language_model.layers.22.mlp.gate_proj.weight",
345
+ "model.language_model.layers.22.mlp.up_proj.weight",
346
+ "model.language_model.layers.22.self_attn.k_proj.weight",
347
+ "model.language_model.layers.22.self_attn.o_proj.weight",
348
+ "model.language_model.layers.22.self_attn.q_proj.weight",
349
+ "model.language_model.layers.22.self_attn.v_proj.weight",
350
+ "model.language_model.layers.23.mlp.down_proj.weight",
351
+ "model.language_model.layers.23.mlp.gate_proj.weight",
352
+ "model.language_model.layers.23.mlp.up_proj.weight",
353
+ "model.language_model.layers.23.self_attn.k_proj.weight",
354
+ "model.language_model.layers.23.self_attn.o_proj.weight",
355
+ "model.language_model.layers.23.self_attn.q_proj.weight",
356
+ "model.language_model.layers.23.self_attn.v_proj.weight",
357
+ "model.language_model.layers.24.mlp.down_proj.weight",
358
+ "model.language_model.layers.24.mlp.gate_proj.weight",
359
+ "model.language_model.layers.24.mlp.up_proj.weight",
360
+ "model.language_model.layers.24.self_attn.k_proj.weight",
361
+ "model.language_model.layers.24.self_attn.o_proj.weight",
362
+ "model.language_model.layers.24.self_attn.q_proj.weight",
363
+ "model.language_model.layers.24.self_attn.v_proj.weight",
364
+ "model.language_model.layers.25.mlp.down_proj.weight",
365
+ "model.language_model.layers.25.mlp.gate_proj.weight",
366
+ "model.language_model.layers.25.mlp.up_proj.weight",
367
+ "model.language_model.layers.25.self_attn.k_proj.weight",
368
+ "model.language_model.layers.25.self_attn.o_proj.weight",
369
+ "model.language_model.layers.25.self_attn.q_proj.weight",
370
+ "model.language_model.layers.25.self_attn.v_proj.weight",
371
+ "model.language_model.layers.26.mlp.down_proj.weight",
372
+ "model.language_model.layers.26.mlp.gate_proj.weight",
373
+ "model.language_model.layers.26.mlp.up_proj.weight",
374
+ "model.language_model.layers.26.self_attn.k_proj.weight",
375
+ "model.language_model.layers.26.self_attn.o_proj.weight",
376
+ "model.language_model.layers.26.self_attn.q_proj.weight",
377
+ "model.language_model.layers.26.self_attn.v_proj.weight",
378
+ "model.language_model.layers.27.mlp.down_proj.weight",
379
+ "model.language_model.layers.27.mlp.gate_proj.weight",
380
+ "model.language_model.layers.27.mlp.up_proj.weight",
381
+ "model.language_model.layers.27.self_attn.k_proj.weight",
382
+ "model.language_model.layers.27.self_attn.o_proj.weight",
383
+ "model.language_model.layers.27.self_attn.q_proj.weight",
384
+ "model.language_model.layers.27.self_attn.v_proj.weight",
385
+ "model.language_model.layers.28.mlp.gate_proj.weight",
386
+ "model.language_model.layers.28.self_attn.k_proj.weight",
387
+ "model.language_model.layers.28.self_attn.o_proj.weight",
388
+ "model.language_model.layers.28.self_attn.q_proj.weight",
389
+ "model.language_model.layers.28.self_attn.v_proj.weight",
390
+ "model.language_model.layers.28.mlp.down_proj.weight",
391
+ "model.language_model.layers.28.mlp.up_proj.weight",
392
+ "model.language_model.layers.29.mlp.down_proj.weight",
393
+ "model.language_model.layers.29.mlp.gate_proj.weight",
394
+ "model.language_model.layers.29.mlp.up_proj.weight",
395
+ "model.language_model.layers.29.self_attn.k_proj.weight",
396
+ "model.language_model.layers.29.self_attn.o_proj.weight",
397
+ "model.language_model.layers.29.self_attn.q_proj.weight",
398
+ "model.language_model.layers.29.self_attn.v_proj.weight",
399
+ "model.language_model.layers.30.mlp.down_proj.weight",
400
+ "model.language_model.layers.30.mlp.gate_proj.weight",
401
+ "model.language_model.layers.30.mlp.up_proj.weight",
402
+ "model.language_model.layers.30.self_attn.k_proj.weight",
403
+ "model.language_model.layers.30.self_attn.o_proj.weight",
404
+ "model.language_model.layers.30.self_attn.q_proj.weight",
405
+ "model.language_model.layers.30.self_attn.v_proj.weight",
406
+ "model.language_model.layers.31.mlp.down_proj.weight",
407
+ "model.language_model.layers.31.mlp.gate_proj.weight",
408
+ "model.language_model.layers.31.mlp.up_proj.weight",
409
+ "model.language_model.layers.31.self_attn.k_proj.weight",
410
+ "model.language_model.layers.31.self_attn.o_proj.weight",
411
+ "model.language_model.layers.31.self_attn.q_proj.weight",
412
+ "model.language_model.layers.31.self_attn.v_proj.weight",
413
+ "model.language_model.layers.32.mlp.down_proj.weight",
414
+ "model.language_model.layers.32.mlp.gate_proj.weight",
415
+ "model.language_model.layers.32.mlp.up_proj.weight",
416
+ "model.language_model.layers.32.self_attn.k_proj.weight",
417
+ "model.language_model.layers.32.self_attn.o_proj.weight",
418
+ "model.language_model.layers.32.self_attn.q_proj.weight",
419
+ "model.language_model.layers.32.self_attn.v_proj.weight",
420
+ "model.language_model.layers.33.mlp.down_proj.weight",
421
+ "model.language_model.layers.33.mlp.gate_proj.weight",
422
+ "model.language_model.layers.33.mlp.up_proj.weight",
423
+ "model.language_model.layers.33.self_attn.k_proj.weight",
424
+ "model.language_model.layers.33.self_attn.o_proj.weight",
425
+ "model.language_model.layers.33.self_attn.q_proj.weight",
426
+ "model.language_model.layers.33.self_attn.v_proj.weight",
427
+ "model.language_model.layers.34.mlp.down_proj.weight",
428
+ "model.language_model.layers.34.mlp.gate_proj.weight",
429
+ "model.language_model.layers.34.mlp.up_proj.weight",
430
+ "model.language_model.layers.34.self_attn.k_proj.weight",
431
+ "model.language_model.layers.34.self_attn.o_proj.weight",
432
+ "model.language_model.layers.34.self_attn.q_proj.weight",
433
+ "model.language_model.layers.34.self_attn.v_proj.weight",
434
+ "model.language_model.layers.35.mlp.down_proj.weight",
435
+ "model.language_model.layers.35.mlp.gate_proj.weight",
436
+ "model.language_model.layers.35.mlp.up_proj.weight",
437
+ "model.language_model.layers.35.self_attn.k_proj.weight",
438
+ "model.language_model.layers.35.self_attn.o_proj.weight",
439
+ "model.language_model.layers.35.self_attn.q_proj.weight",
440
+ "model.language_model.layers.35.self_attn.v_proj.weight",
441
+ "model.language_model.layers.36.mlp.down_proj.weight",
442
+ "model.language_model.layers.36.mlp.gate_proj.weight",
443
+ "model.language_model.layers.36.mlp.up_proj.weight",
444
+ "model.language_model.layers.36.self_attn.k_proj.weight",
445
+ "model.language_model.layers.36.self_attn.o_proj.weight",
446
+ "model.language_model.layers.36.self_attn.q_proj.weight",
447
+ "model.language_model.layers.36.self_attn.v_proj.weight",
448
+ "model.language_model.layers.37.mlp.down_proj.weight",
449
+ "model.language_model.layers.37.mlp.gate_proj.weight",
450
+ "model.language_model.layers.37.mlp.up_proj.weight",
451
+ "model.language_model.layers.37.self_attn.k_proj.weight",
452
+ "model.language_model.layers.37.self_attn.o_proj.weight",
453
+ "model.language_model.layers.37.self_attn.q_proj.weight",
454
+ "model.language_model.layers.37.self_attn.v_proj.weight",
455
+ "model.language_model.layers.38.mlp.down_proj.weight",
456
+ "model.language_model.layers.38.mlp.gate_proj.weight",
457
+ "model.language_model.layers.38.mlp.up_proj.weight",
458
+ "model.language_model.layers.38.self_attn.k_proj.weight",
459
+ "model.language_model.layers.38.self_attn.o_proj.weight",
460
+ "model.language_model.layers.38.self_attn.q_proj.weight",
461
+ "model.language_model.layers.38.self_attn.v_proj.weight",
462
+ "model.language_model.layers.39.mlp.gate_proj.weight",
463
+ "model.language_model.layers.39.self_attn.k_proj.weight",
464
+ "model.language_model.layers.39.self_attn.o_proj.weight",
465
+ "model.language_model.layers.39.self_attn.q_proj.weight",
466
+ "model.language_model.layers.39.self_attn.v_proj.weight",
467
+ "model.language_model.layers.39.mlp.down_proj.weight",
468
+ "model.language_model.layers.39.mlp.up_proj.weight",
469
+ "model.language_model.layers.40.mlp.down_proj.weight",
470
+ "model.language_model.layers.40.mlp.gate_proj.weight",
471
+ "model.language_model.layers.40.mlp.up_proj.weight",
472
+ "model.language_model.layers.40.self_attn.k_proj.weight",
473
+ "model.language_model.layers.40.self_attn.o_proj.weight",
474
+ "model.language_model.layers.40.self_attn.q_proj.weight",
475
+ "model.language_model.layers.40.self_attn.v_proj.weight",
476
+ "model.language_model.layers.41.mlp.down_proj.weight",
477
+ "model.language_model.layers.41.mlp.gate_proj.weight",
478
+ "model.language_model.layers.41.mlp.up_proj.weight",
479
+ "model.language_model.layers.41.self_attn.k_proj.weight",
480
+ "model.language_model.layers.41.self_attn.o_proj.weight",
481
+ "model.language_model.layers.41.self_attn.q_proj.weight",
482
+ "model.language_model.layers.41.self_attn.v_proj.weight",
483
+ "model.language_model.layers.42.mlp.down_proj.weight",
484
+ "model.language_model.layers.42.mlp.gate_proj.weight",
485
+ "model.language_model.layers.42.mlp.up_proj.weight",
486
+ "model.language_model.layers.42.self_attn.k_proj.weight",
487
+ "model.language_model.layers.42.self_attn.o_proj.weight",
488
+ "model.language_model.layers.42.self_attn.q_proj.weight",
489
+ "model.language_model.layers.42.self_attn.v_proj.weight",
490
+ "model.language_model.layers.43.mlp.down_proj.weight",
491
+ "model.language_model.layers.43.mlp.gate_proj.weight",
492
+ "model.language_model.layers.43.mlp.up_proj.weight",
493
+ "model.language_model.layers.43.self_attn.k_proj.weight",
494
+ "model.language_model.layers.43.self_attn.o_proj.weight",
495
+ "model.language_model.layers.43.self_attn.q_proj.weight",
496
+ "model.language_model.layers.43.self_attn.v_proj.weight",
497
+ "model.language_model.layers.44.mlp.down_proj.weight",
498
+ "model.language_model.layers.44.mlp.gate_proj.weight",
499
+ "model.language_model.layers.44.mlp.up_proj.weight",
500
+ "model.language_model.layers.44.self_attn.k_proj.weight",
501
+ "model.language_model.layers.44.self_attn.o_proj.weight",
502
+ "model.language_model.layers.44.self_attn.q_proj.weight",
503
+ "model.language_model.layers.44.self_attn.v_proj.weight",
504
+ "model.language_model.layers.45.mlp.down_proj.weight",
505
+ "model.language_model.layers.45.mlp.gate_proj.weight",
506
+ "model.language_model.layers.45.mlp.up_proj.weight",
507
+ "model.language_model.layers.45.self_attn.k_proj.weight",
508
+ "model.language_model.layers.45.self_attn.o_proj.weight",
509
+ "model.language_model.layers.45.self_attn.q_proj.weight",
510
+ "model.language_model.layers.45.self_attn.v_proj.weight",
511
+ "model.language_model.layers.46.mlp.down_proj.weight",
512
+ "model.language_model.layers.46.mlp.gate_proj.weight",
513
+ "model.language_model.layers.46.mlp.up_proj.weight",
514
+ "model.language_model.layers.46.self_attn.k_proj.weight",
515
+ "model.language_model.layers.46.self_attn.o_proj.weight",
516
+ "model.language_model.layers.46.self_attn.q_proj.weight",
517
+ "model.language_model.layers.46.self_attn.v_proj.weight",
518
+ "model.language_model.layers.47.mlp.down_proj.weight",
519
+ "model.language_model.layers.47.mlp.gate_proj.weight",
520
+ "model.language_model.layers.47.mlp.up_proj.weight",
521
+ "model.language_model.layers.47.self_attn.k_proj.weight",
522
+ "model.language_model.layers.47.self_attn.o_proj.weight",
523
+ "model.language_model.layers.47.self_attn.q_proj.weight",
524
+ "model.language_model.layers.47.self_attn.v_proj.weight"
525
+ ]
526
+ },
527
+ "modules_to_not_convert": [
528
+ ".condition_embedder",
529
+ ".txt_out",
530
+ "lm_head.weight",
531
+ ".vid_out",
532
+ ".emb_out",
533
+ ".emb_in",
534
+ "model.language_model.embed_tokens",
535
+ "lm_head",
536
+ ".img_out",
537
+ ".t_embedder",
538
+ ".norm_out",
539
+ ".txt_in",
540
+ "time_text_embed",
541
+ ".context_embedder",
542
+ ".y_embedder",
543
+ "patch_emb",
544
+ "wte",
545
+ ".final_layer",
546
+ ".img_in",
547
+ ".x_embedder",
548
+ "multi_modal_projector",
549
+ "patch_embed",
550
+ ".time_embed",
551
+ "model.language_model.norm",
552
+ ".vid_in",
553
+ ".proj_out",
554
+ "patch_embedding"
555
+ ],
556
+ "non_blocking": false,
557
+ "quant_conv": false,
558
+ "quant_method": "sdnq",
559
+ "quantization_device": null,
560
+ "quantized_matmul_dtype": "int8",
561
+ "return_device": null,
562
+ "sdnq_version": "0.1.4",
563
+ "svd_rank": 32,
564
+ "svd_steps": 8,
565
+ "use_dynamic_quantization": true,
566
+ "use_grad_ckpt": true,
567
+ "use_quantized_matmul": false,
568
+ "use_quantized_matmul_conv": false,
569
+ "use_static_quantization": true,
570
+ "use_stochastic_rounding": false,
571
+ "use_svd": false,
572
+ "weights_dtype": "int4"
573
+ },
574
+ "text_config": {
575
+ "_sliding_window_pattern": 6,
576
+ "attention_bias": false,
577
+ "attention_dropout": 0.0,
578
+ "attn_logit_softcapping": null,
579
+ "cache_implementation": "hybrid",
580
+ "dtype": "bfloat16",
581
+ "final_logit_softcapping": null,
582
+ "head_dim": 256,
583
+ "hidden_activation": "gelu_pytorch_tanh",
584
+ "hidden_size": 3840,
585
+ "initializer_range": 0.02,
586
+ "intermediate_size": 15360,
587
+ "layer_types": [
588
+ "sliding_attention",
589
+ "sliding_attention",
590
+ "sliding_attention",
591
+ "sliding_attention",
592
+ "sliding_attention",
593
+ "full_attention",
594
+ "sliding_attention",
595
+ "sliding_attention",
596
+ "sliding_attention",
597
+ "sliding_attention",
598
+ "sliding_attention",
599
+ "full_attention",
600
+ "sliding_attention",
601
+ "sliding_attention",
602
+ "sliding_attention",
603
+ "sliding_attention",
604
+ "sliding_attention",
605
+ "full_attention",
606
+ "sliding_attention",
607
+ "sliding_attention",
608
+ "sliding_attention",
609
+ "sliding_attention",
610
+ "sliding_attention",
611
+ "full_attention",
612
+ "sliding_attention",
613
+ "sliding_attention",
614
+ "sliding_attention",
615
+ "sliding_attention",
616
+ "sliding_attention",
617
+ "full_attention",
618
+ "sliding_attention",
619
+ "sliding_attention",
620
+ "sliding_attention",
621
+ "sliding_attention",
622
+ "sliding_attention",
623
+ "full_attention",
624
+ "sliding_attention",
625
+ "sliding_attention",
626
+ "sliding_attention",
627
+ "sliding_attention",
628
+ "sliding_attention",
629
+ "full_attention",
630
+ "sliding_attention",
631
+ "sliding_attention",
632
+ "sliding_attention",
633
+ "sliding_attention",
634
+ "sliding_attention",
635
+ "full_attention"
636
+ ],
637
+ "max_position_embeddings": 131072,
638
+ "model_type": "gemma3_text",
639
+ "num_attention_heads": 16,
640
+ "num_hidden_layers": 48,
641
+ "num_key_value_heads": 8,
642
+ "query_pre_attn_scalar": 256,
643
+ "rms_norm_eps": 1e-06,
644
+ "rope_local_base_freq": 10000,
645
+ "rope_scaling": {
646
+ "factor": 8.0,
647
+ "rope_type": "linear"
648
+ },
649
+ "rope_theta": 1000000,
650
+ "sliding_window": 1024,
651
+ "sliding_window_pattern": 6,
652
+ "use_bidirectional_attention": false,
653
+ "use_cache": true,
654
+ "vocab_size": 262208
655
+ },
656
+ "transformers_version": "4.57.3",
657
+ "vision_config": {
658
+ "attention_dropout": 0.0,
659
+ "dtype": "bfloat16",
660
+ "hidden_act": "gelu_pytorch_tanh",
661
+ "hidden_size": 1152,
662
+ "image_size": 896,
663
+ "intermediate_size": 4304,
664
+ "layer_norm_eps": 1e-06,
665
+ "model_type": "siglip_vision_model",
666
+ "num_attention_heads": 16,
667
+ "num_channels": 3,
668
+ "num_hidden_layers": 27,
669
+ "patch_size": 14,
670
+ "vision_use_head": false
671
+ }
672
+ }
text_encoder/generation_config.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cache_implementation": "hybrid",
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 1,
6
+ 106
7
+ ],
8
+ "top_k": 64,
9
+ "top_p": 0.95,
10
+ "transformers_version": "4.57.3"
11
+ }
text_encoder/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dadabab660c760cf65814abc3dba1968dc894aac8768df4f109fbbd5dccf3038
3
+ size 4985945456
text_encoder/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c1d778b16ba9c36ee029c1d5bc0a5282c67dd06c48bac67011ae5d3c090d8e0
3
+ size 3345258264
text_encoder/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
text_encoder/quantization_config.json ADDED
@@ -0,0 +1,558 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_skip_keys": false,
3
+ "dequantize_fp32": false,
4
+ "dynamic_loss_threshold": 0.01,
5
+ "group_size": 0,
6
+ "is_integer": true,
7
+ "is_training": false,
8
+ "modules_dtype_dict": {
9
+ "int5": [
10
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight",
11
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight",
12
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight",
13
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight",
14
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight",
15
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight",
16
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight",
17
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight",
18
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight",
19
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight",
20
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight",
21
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight",
22
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight",
23
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight",
24
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight",
25
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight",
26
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight",
27
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight",
28
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight",
29
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight",
30
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight",
31
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight",
32
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight",
33
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight",
34
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight",
35
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight",
36
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight",
37
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight",
38
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight",
39
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight",
40
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight",
41
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight",
42
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight",
43
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc2.weight",
44
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight",
45
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight",
46
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight",
47
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight",
48
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight",
49
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight",
50
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight",
51
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight",
52
+ "model.language_model.layers.17.self_attn.k_proj.weight"
53
+ ],
54
+ "uint4": [
55
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight",
56
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight",
57
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight",
58
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight",
59
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight",
60
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight",
61
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight",
62
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight",
63
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight",
64
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight",
65
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight",
66
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight",
67
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight",
68
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight",
69
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight",
70
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight",
71
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight",
72
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight",
73
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight",
74
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight",
75
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight",
76
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight",
77
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight",
78
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight",
79
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight",
80
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight",
81
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight",
82
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight",
83
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight",
84
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight",
85
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight",
86
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight",
87
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight",
88
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight",
89
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight",
90
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight",
91
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight",
92
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight",
93
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight",
94
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight",
95
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight",
96
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight",
97
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight",
98
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight",
99
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight",
100
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight",
101
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight",
102
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight",
103
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight",
104
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight",
105
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight",
106
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight",
107
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight",
108
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight",
109
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight",
110
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight",
111
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight",
112
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight",
113
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight",
114
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight",
115
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight",
116
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight",
117
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight",
118
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight",
119
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight",
120
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight",
121
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight",
122
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight",
123
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight",
124
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight",
125
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight",
126
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight",
127
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight",
128
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight",
129
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight",
130
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight",
131
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight",
132
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight",
133
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight",
134
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight",
135
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight",
136
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc1.weight",
137
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj.weight",
138
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj.weight",
139
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj.weight",
140
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj.weight",
141
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight",
142
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight",
143
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight",
144
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight",
145
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight",
146
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight",
147
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight",
148
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight",
149
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight",
150
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight",
151
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight",
152
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight",
153
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight",
154
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight",
155
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight",
156
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight",
157
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight",
158
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight",
159
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight",
160
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight",
161
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight",
162
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight",
163
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight",
164
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight",
165
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight",
166
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight",
167
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight",
168
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight",
169
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight",
170
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight",
171
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight",
172
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight",
173
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight",
174
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight",
175
+ "model.language_model.layers.0.mlp.down_proj.weight",
176
+ "model.language_model.layers.0.mlp.gate_proj.weight",
177
+ "model.language_model.layers.0.mlp.up_proj.weight",
178
+ "model.language_model.layers.0.self_attn.k_proj.weight",
179
+ "model.language_model.layers.0.self_attn.o_proj.weight",
180
+ "model.language_model.layers.0.self_attn.q_proj.weight",
181
+ "model.language_model.layers.0.self_attn.v_proj.weight",
182
+ "model.language_model.layers.1.self_attn.q_proj.weight",
183
+ "model.language_model.layers.1.mlp.down_proj.weight",
184
+ "model.language_model.layers.1.mlp.gate_proj.weight",
185
+ "model.language_model.layers.1.mlp.up_proj.weight",
186
+ "model.language_model.layers.1.self_attn.k_proj.weight",
187
+ "model.language_model.layers.1.self_attn.o_proj.weight",
188
+ "model.language_model.layers.1.self_attn.v_proj.weight",
189
+ "model.language_model.layers.2.mlp.down_proj.weight",
190
+ "model.language_model.layers.2.mlp.gate_proj.weight",
191
+ "model.language_model.layers.2.mlp.up_proj.weight",
192
+ "model.language_model.layers.2.self_attn.k_proj.weight",
193
+ "model.language_model.layers.2.self_attn.o_proj.weight",
194
+ "model.language_model.layers.2.self_attn.q_proj.weight",
195
+ "model.language_model.layers.2.self_attn.v_proj.weight",
196
+ "model.language_model.layers.3.mlp.down_proj.weight",
197
+ "model.language_model.layers.3.mlp.gate_proj.weight",
198
+ "model.language_model.layers.3.mlp.up_proj.weight",
199
+ "model.language_model.layers.3.self_attn.k_proj.weight",
200
+ "model.language_model.layers.3.self_attn.o_proj.weight",
201
+ "model.language_model.layers.3.self_attn.q_proj.weight",
202
+ "model.language_model.layers.3.self_attn.v_proj.weight",
203
+ "model.language_model.layers.4.mlp.down_proj.weight",
204
+ "model.language_model.layers.4.mlp.gate_proj.weight",
205
+ "model.language_model.layers.4.mlp.up_proj.weight",
206
+ "model.language_model.layers.4.self_attn.k_proj.weight",
207
+ "model.language_model.layers.4.self_attn.o_proj.weight",
208
+ "model.language_model.layers.4.self_attn.q_proj.weight",
209
+ "model.language_model.layers.4.self_attn.v_proj.weight",
210
+ "model.language_model.layers.5.mlp.down_proj.weight",
211
+ "model.language_model.layers.5.mlp.gate_proj.weight",
212
+ "model.language_model.layers.5.mlp.up_proj.weight",
213
+ "model.language_model.layers.5.self_attn.k_proj.weight",
214
+ "model.language_model.layers.5.self_attn.o_proj.weight",
215
+ "model.language_model.layers.5.self_attn.q_proj.weight",
216
+ "model.language_model.layers.5.self_attn.v_proj.weight",
217
+ "model.language_model.layers.6.mlp.gate_proj.weight",
218
+ "model.language_model.layers.6.self_attn.k_proj.weight",
219
+ "model.language_model.layers.6.self_attn.o_proj.weight",
220
+ "model.language_model.layers.6.self_attn.q_proj.weight",
221
+ "model.language_model.layers.6.self_attn.v_proj.weight",
222
+ "model.language_model.layers.10.mlp.down_proj.weight",
223
+ "model.language_model.layers.10.mlp.gate_proj.weight",
224
+ "model.language_model.layers.10.mlp.up_proj.weight",
225
+ "model.language_model.layers.10.self_attn.k_proj.weight",
226
+ "model.language_model.layers.10.self_attn.o_proj.weight",
227
+ "model.language_model.layers.10.self_attn.q_proj.weight",
228
+ "model.language_model.layers.10.self_attn.v_proj.weight",
229
+ "model.language_model.layers.11.mlp.down_proj.weight",
230
+ "model.language_model.layers.11.mlp.gate_proj.weight",
231
+ "model.language_model.layers.11.mlp.up_proj.weight",
232
+ "model.language_model.layers.11.self_attn.k_proj.weight",
233
+ "model.language_model.layers.11.self_attn.o_proj.weight",
234
+ "model.language_model.layers.11.self_attn.q_proj.weight",
235
+ "model.language_model.layers.11.self_attn.v_proj.weight",
236
+ "model.language_model.layers.6.mlp.down_proj.weight",
237
+ "model.language_model.layers.6.mlp.up_proj.weight",
238
+ "model.language_model.layers.7.mlp.down_proj.weight",
239
+ "model.language_model.layers.7.mlp.gate_proj.weight",
240
+ "model.language_model.layers.7.mlp.up_proj.weight",
241
+ "model.language_model.layers.7.self_attn.k_proj.weight",
242
+ "model.language_model.layers.7.self_attn.o_proj.weight",
243
+ "model.language_model.layers.7.self_attn.q_proj.weight",
244
+ "model.language_model.layers.7.self_attn.v_proj.weight",
245
+ "model.language_model.layers.8.mlp.down_proj.weight",
246
+ "model.language_model.layers.8.mlp.gate_proj.weight",
247
+ "model.language_model.layers.8.mlp.up_proj.weight",
248
+ "model.language_model.layers.8.self_attn.k_proj.weight",
249
+ "model.language_model.layers.8.self_attn.o_proj.weight",
250
+ "model.language_model.layers.8.self_attn.q_proj.weight",
251
+ "model.language_model.layers.8.self_attn.v_proj.weight",
252
+ "model.language_model.layers.9.mlp.down_proj.weight",
253
+ "model.language_model.layers.9.mlp.gate_proj.weight",
254
+ "model.language_model.layers.9.mlp.up_proj.weight",
255
+ "model.language_model.layers.9.self_attn.k_proj.weight",
256
+ "model.language_model.layers.9.self_attn.o_proj.weight",
257
+ "model.language_model.layers.9.self_attn.q_proj.weight",
258
+ "model.language_model.layers.9.self_attn.v_proj.weight",
259
+ "model.language_model.layers.12.mlp.down_proj.weight",
260
+ "model.language_model.layers.12.mlp.gate_proj.weight",
261
+ "model.language_model.layers.12.mlp.up_proj.weight",
262
+ "model.language_model.layers.12.self_attn.k_proj.weight",
263
+ "model.language_model.layers.12.self_attn.o_proj.weight",
264
+ "model.language_model.layers.12.self_attn.q_proj.weight",
265
+ "model.language_model.layers.12.self_attn.v_proj.weight",
266
+ "model.language_model.layers.13.mlp.down_proj.weight",
267
+ "model.language_model.layers.13.mlp.gate_proj.weight",
268
+ "model.language_model.layers.13.mlp.up_proj.weight",
269
+ "model.language_model.layers.13.self_attn.k_proj.weight",
270
+ "model.language_model.layers.13.self_attn.o_proj.weight",
271
+ "model.language_model.layers.13.self_attn.q_proj.weight",
272
+ "model.language_model.layers.13.self_attn.v_proj.weight",
273
+ "model.language_model.layers.14.mlp.down_proj.weight",
274
+ "model.language_model.layers.14.mlp.gate_proj.weight",
275
+ "model.language_model.layers.14.mlp.up_proj.weight",
276
+ "model.language_model.layers.14.self_attn.k_proj.weight",
277
+ "model.language_model.layers.14.self_attn.o_proj.weight",
278
+ "model.language_model.layers.14.self_attn.q_proj.weight",
279
+ "model.language_model.layers.14.self_attn.v_proj.weight",
280
+ "model.language_model.layers.15.mlp.down_proj.weight",
281
+ "model.language_model.layers.15.mlp.gate_proj.weight",
282
+ "model.language_model.layers.15.mlp.up_proj.weight",
283
+ "model.language_model.layers.15.self_attn.k_proj.weight",
284
+ "model.language_model.layers.15.self_attn.o_proj.weight",
285
+ "model.language_model.layers.15.self_attn.q_proj.weight",
286
+ "model.language_model.layers.15.self_attn.v_proj.weight",
287
+ "model.language_model.layers.16.mlp.down_proj.weight",
288
+ "model.language_model.layers.16.mlp.gate_proj.weight",
289
+ "model.language_model.layers.16.mlp.up_proj.weight",
290
+ "model.language_model.layers.16.self_attn.k_proj.weight",
291
+ "model.language_model.layers.16.self_attn.o_proj.weight",
292
+ "model.language_model.layers.16.self_attn.q_proj.weight",
293
+ "model.language_model.layers.16.self_attn.v_proj.weight",
294
+ "model.language_model.layers.17.mlp.gate_proj.weight",
295
+ "model.language_model.layers.17.self_attn.o_proj.weight",
296
+ "model.language_model.layers.17.self_attn.q_proj.weight",
297
+ "model.language_model.layers.17.self_attn.v_proj.weight",
298
+ "model.language_model.layers.17.mlp.down_proj.weight",
299
+ "model.language_model.layers.17.mlp.up_proj.weight",
300
+ "model.language_model.layers.18.mlp.down_proj.weight",
301
+ "model.language_model.layers.18.mlp.gate_proj.weight",
302
+ "model.language_model.layers.18.mlp.up_proj.weight",
303
+ "model.language_model.layers.18.self_attn.k_proj.weight",
304
+ "model.language_model.layers.18.self_attn.o_proj.weight",
305
+ "model.language_model.layers.18.self_attn.q_proj.weight",
306
+ "model.language_model.layers.18.self_attn.v_proj.weight",
307
+ "model.language_model.layers.19.mlp.down_proj.weight",
308
+ "model.language_model.layers.19.mlp.gate_proj.weight",
309
+ "model.language_model.layers.19.mlp.up_proj.weight",
310
+ "model.language_model.layers.19.self_attn.k_proj.weight",
311
+ "model.language_model.layers.19.self_attn.o_proj.weight",
312
+ "model.language_model.layers.19.self_attn.q_proj.weight",
313
+ "model.language_model.layers.19.self_attn.v_proj.weight",
314
+ "model.language_model.layers.20.mlp.down_proj.weight",
315
+ "model.language_model.layers.20.mlp.gate_proj.weight",
316
+ "model.language_model.layers.20.mlp.up_proj.weight",
317
+ "model.language_model.layers.20.self_attn.k_proj.weight",
318
+ "model.language_model.layers.20.self_attn.o_proj.weight",
319
+ "model.language_model.layers.20.self_attn.q_proj.weight",
320
+ "model.language_model.layers.20.self_attn.v_proj.weight",
321
+ "model.language_model.layers.21.mlp.down_proj.weight",
322
+ "model.language_model.layers.21.mlp.gate_proj.weight",
323
+ "model.language_model.layers.21.mlp.up_proj.weight",
324
+ "model.language_model.layers.21.self_attn.k_proj.weight",
325
+ "model.language_model.layers.21.self_attn.o_proj.weight",
326
+ "model.language_model.layers.21.self_attn.q_proj.weight",
327
+ "model.language_model.layers.21.self_attn.v_proj.weight",
328
+ "model.language_model.layers.22.mlp.down_proj.weight",
329
+ "model.language_model.layers.22.mlp.gate_proj.weight",
330
+ "model.language_model.layers.22.mlp.up_proj.weight",
331
+ "model.language_model.layers.22.self_attn.k_proj.weight",
332
+ "model.language_model.layers.22.self_attn.o_proj.weight",
333
+ "model.language_model.layers.22.self_attn.q_proj.weight",
334
+ "model.language_model.layers.22.self_attn.v_proj.weight",
335
+ "model.language_model.layers.23.mlp.down_proj.weight",
336
+ "model.language_model.layers.23.mlp.gate_proj.weight",
337
+ "model.language_model.layers.23.mlp.up_proj.weight",
338
+ "model.language_model.layers.23.self_attn.k_proj.weight",
339
+ "model.language_model.layers.23.self_attn.o_proj.weight",
340
+ "model.language_model.layers.23.self_attn.q_proj.weight",
341
+ "model.language_model.layers.23.self_attn.v_proj.weight",
342
+ "model.language_model.layers.24.mlp.down_proj.weight",
343
+ "model.language_model.layers.24.mlp.gate_proj.weight",
344
+ "model.language_model.layers.24.mlp.up_proj.weight",
345
+ "model.language_model.layers.24.self_attn.k_proj.weight",
346
+ "model.language_model.layers.24.self_attn.o_proj.weight",
347
+ "model.language_model.layers.24.self_attn.q_proj.weight",
348
+ "model.language_model.layers.24.self_attn.v_proj.weight",
349
+ "model.language_model.layers.25.mlp.down_proj.weight",
350
+ "model.language_model.layers.25.mlp.gate_proj.weight",
351
+ "model.language_model.layers.25.mlp.up_proj.weight",
352
+ "model.language_model.layers.25.self_attn.k_proj.weight",
353
+ "model.language_model.layers.25.self_attn.o_proj.weight",
354
+ "model.language_model.layers.25.self_attn.q_proj.weight",
355
+ "model.language_model.layers.25.self_attn.v_proj.weight",
356
+ "model.language_model.layers.26.mlp.down_proj.weight",
357
+ "model.language_model.layers.26.mlp.gate_proj.weight",
358
+ "model.language_model.layers.26.mlp.up_proj.weight",
359
+ "model.language_model.layers.26.self_attn.k_proj.weight",
360
+ "model.language_model.layers.26.self_attn.o_proj.weight",
361
+ "model.language_model.layers.26.self_attn.q_proj.weight",
362
+ "model.language_model.layers.26.self_attn.v_proj.weight",
363
+ "model.language_model.layers.27.mlp.down_proj.weight",
364
+ "model.language_model.layers.27.mlp.gate_proj.weight",
365
+ "model.language_model.layers.27.mlp.up_proj.weight",
366
+ "model.language_model.layers.27.self_attn.k_proj.weight",
367
+ "model.language_model.layers.27.self_attn.o_proj.weight",
368
+ "model.language_model.layers.27.self_attn.q_proj.weight",
369
+ "model.language_model.layers.27.self_attn.v_proj.weight",
370
+ "model.language_model.layers.28.mlp.gate_proj.weight",
371
+ "model.language_model.layers.28.self_attn.k_proj.weight",
372
+ "model.language_model.layers.28.self_attn.o_proj.weight",
373
+ "model.language_model.layers.28.self_attn.q_proj.weight",
374
+ "model.language_model.layers.28.self_attn.v_proj.weight",
375
+ "model.language_model.layers.28.mlp.down_proj.weight",
376
+ "model.language_model.layers.28.mlp.up_proj.weight",
377
+ "model.language_model.layers.29.mlp.down_proj.weight",
378
+ "model.language_model.layers.29.mlp.gate_proj.weight",
379
+ "model.language_model.layers.29.mlp.up_proj.weight",
380
+ "model.language_model.layers.29.self_attn.k_proj.weight",
381
+ "model.language_model.layers.29.self_attn.o_proj.weight",
382
+ "model.language_model.layers.29.self_attn.q_proj.weight",
383
+ "model.language_model.layers.29.self_attn.v_proj.weight",
384
+ "model.language_model.layers.30.mlp.down_proj.weight",
385
+ "model.language_model.layers.30.mlp.gate_proj.weight",
386
+ "model.language_model.layers.30.mlp.up_proj.weight",
387
+ "model.language_model.layers.30.self_attn.k_proj.weight",
388
+ "model.language_model.layers.30.self_attn.o_proj.weight",
389
+ "model.language_model.layers.30.self_attn.q_proj.weight",
390
+ "model.language_model.layers.30.self_attn.v_proj.weight",
391
+ "model.language_model.layers.31.mlp.down_proj.weight",
392
+ "model.language_model.layers.31.mlp.gate_proj.weight",
393
+ "model.language_model.layers.31.mlp.up_proj.weight",
394
+ "model.language_model.layers.31.self_attn.k_proj.weight",
395
+ "model.language_model.layers.31.self_attn.o_proj.weight",
396
+ "model.language_model.layers.31.self_attn.q_proj.weight",
397
+ "model.language_model.layers.31.self_attn.v_proj.weight",
398
+ "model.language_model.layers.32.mlp.down_proj.weight",
399
+ "model.language_model.layers.32.mlp.gate_proj.weight",
400
+ "model.language_model.layers.32.mlp.up_proj.weight",
401
+ "model.language_model.layers.32.self_attn.k_proj.weight",
402
+ "model.language_model.layers.32.self_attn.o_proj.weight",
403
+ "model.language_model.layers.32.self_attn.q_proj.weight",
404
+ "model.language_model.layers.32.self_attn.v_proj.weight",
405
+ "model.language_model.layers.33.mlp.down_proj.weight",
406
+ "model.language_model.layers.33.mlp.gate_proj.weight",
407
+ "model.language_model.layers.33.mlp.up_proj.weight",
408
+ "model.language_model.layers.33.self_attn.k_proj.weight",
409
+ "model.language_model.layers.33.self_attn.o_proj.weight",
410
+ "model.language_model.layers.33.self_attn.q_proj.weight",
411
+ "model.language_model.layers.33.self_attn.v_proj.weight",
412
+ "model.language_model.layers.34.mlp.down_proj.weight",
413
+ "model.language_model.layers.34.mlp.gate_proj.weight",
414
+ "model.language_model.layers.34.mlp.up_proj.weight",
415
+ "model.language_model.layers.34.self_attn.k_proj.weight",
416
+ "model.language_model.layers.34.self_attn.o_proj.weight",
417
+ "model.language_model.layers.34.self_attn.q_proj.weight",
418
+ "model.language_model.layers.34.self_attn.v_proj.weight",
419
+ "model.language_model.layers.35.mlp.down_proj.weight",
420
+ "model.language_model.layers.35.mlp.gate_proj.weight",
421
+ "model.language_model.layers.35.mlp.up_proj.weight",
422
+ "model.language_model.layers.35.self_attn.k_proj.weight",
423
+ "model.language_model.layers.35.self_attn.o_proj.weight",
424
+ "model.language_model.layers.35.self_attn.q_proj.weight",
425
+ "model.language_model.layers.35.self_attn.v_proj.weight",
426
+ "model.language_model.layers.36.mlp.down_proj.weight",
427
+ "model.language_model.layers.36.mlp.gate_proj.weight",
428
+ "model.language_model.layers.36.mlp.up_proj.weight",
429
+ "model.language_model.layers.36.self_attn.k_proj.weight",
430
+ "model.language_model.layers.36.self_attn.o_proj.weight",
431
+ "model.language_model.layers.36.self_attn.q_proj.weight",
432
+ "model.language_model.layers.36.self_attn.v_proj.weight",
433
+ "model.language_model.layers.37.mlp.down_proj.weight",
434
+ "model.language_model.layers.37.mlp.gate_proj.weight",
435
+ "model.language_model.layers.37.mlp.up_proj.weight",
436
+ "model.language_model.layers.37.self_attn.k_proj.weight",
437
+ "model.language_model.layers.37.self_attn.o_proj.weight",
438
+ "model.language_model.layers.37.self_attn.q_proj.weight",
439
+ "model.language_model.layers.37.self_attn.v_proj.weight",
440
+ "model.language_model.layers.38.mlp.down_proj.weight",
441
+ "model.language_model.layers.38.mlp.gate_proj.weight",
442
+ "model.language_model.layers.38.mlp.up_proj.weight",
443
+ "model.language_model.layers.38.self_attn.k_proj.weight",
444
+ "model.language_model.layers.38.self_attn.o_proj.weight",
445
+ "model.language_model.layers.38.self_attn.q_proj.weight",
446
+ "model.language_model.layers.38.self_attn.v_proj.weight",
447
+ "model.language_model.layers.39.mlp.gate_proj.weight",
448
+ "model.language_model.layers.39.self_attn.k_proj.weight",
449
+ "model.language_model.layers.39.self_attn.o_proj.weight",
450
+ "model.language_model.layers.39.self_attn.q_proj.weight",
451
+ "model.language_model.layers.39.self_attn.v_proj.weight",
452
+ "model.language_model.layers.39.mlp.down_proj.weight",
453
+ "model.language_model.layers.39.mlp.up_proj.weight",
454
+ "model.language_model.layers.40.mlp.down_proj.weight",
455
+ "model.language_model.layers.40.mlp.gate_proj.weight",
456
+ "model.language_model.layers.40.mlp.up_proj.weight",
457
+ "model.language_model.layers.40.self_attn.k_proj.weight",
458
+ "model.language_model.layers.40.self_attn.o_proj.weight",
459
+ "model.language_model.layers.40.self_attn.q_proj.weight",
460
+ "model.language_model.layers.40.self_attn.v_proj.weight",
461
+ "model.language_model.layers.41.mlp.down_proj.weight",
462
+ "model.language_model.layers.41.mlp.gate_proj.weight",
463
+ "model.language_model.layers.41.mlp.up_proj.weight",
464
+ "model.language_model.layers.41.self_attn.k_proj.weight",
465
+ "model.language_model.layers.41.self_attn.o_proj.weight",
466
+ "model.language_model.layers.41.self_attn.q_proj.weight",
467
+ "model.language_model.layers.41.self_attn.v_proj.weight",
468
+ "model.language_model.layers.42.mlp.down_proj.weight",
469
+ "model.language_model.layers.42.mlp.gate_proj.weight",
470
+ "model.language_model.layers.42.mlp.up_proj.weight",
471
+ "model.language_model.layers.42.self_attn.k_proj.weight",
472
+ "model.language_model.layers.42.self_attn.o_proj.weight",
473
+ "model.language_model.layers.42.self_attn.q_proj.weight",
474
+ "model.language_model.layers.42.self_attn.v_proj.weight",
475
+ "model.language_model.layers.43.mlp.down_proj.weight",
476
+ "model.language_model.layers.43.mlp.gate_proj.weight",
477
+ "model.language_model.layers.43.mlp.up_proj.weight",
478
+ "model.language_model.layers.43.self_attn.k_proj.weight",
479
+ "model.language_model.layers.43.self_attn.o_proj.weight",
480
+ "model.language_model.layers.43.self_attn.q_proj.weight",
481
+ "model.language_model.layers.43.self_attn.v_proj.weight",
482
+ "model.language_model.layers.44.mlp.down_proj.weight",
483
+ "model.language_model.layers.44.mlp.gate_proj.weight",
484
+ "model.language_model.layers.44.mlp.up_proj.weight",
485
+ "model.language_model.layers.44.self_attn.k_proj.weight",
486
+ "model.language_model.layers.44.self_attn.o_proj.weight",
487
+ "model.language_model.layers.44.self_attn.q_proj.weight",
488
+ "model.language_model.layers.44.self_attn.v_proj.weight",
489
+ "model.language_model.layers.45.mlp.down_proj.weight",
490
+ "model.language_model.layers.45.mlp.gate_proj.weight",
491
+ "model.language_model.layers.45.mlp.up_proj.weight",
492
+ "model.language_model.layers.45.self_attn.k_proj.weight",
493
+ "model.language_model.layers.45.self_attn.o_proj.weight",
494
+ "model.language_model.layers.45.self_attn.q_proj.weight",
495
+ "model.language_model.layers.45.self_attn.v_proj.weight",
496
+ "model.language_model.layers.46.mlp.down_proj.weight",
497
+ "model.language_model.layers.46.mlp.gate_proj.weight",
498
+ "model.language_model.layers.46.mlp.up_proj.weight",
499
+ "model.language_model.layers.46.self_attn.k_proj.weight",
500
+ "model.language_model.layers.46.self_attn.o_proj.weight",
501
+ "model.language_model.layers.46.self_attn.q_proj.weight",
502
+ "model.language_model.layers.46.self_attn.v_proj.weight",
503
+ "model.language_model.layers.47.mlp.down_proj.weight",
504
+ "model.language_model.layers.47.mlp.gate_proj.weight",
505
+ "model.language_model.layers.47.mlp.up_proj.weight",
506
+ "model.language_model.layers.47.self_attn.k_proj.weight",
507
+ "model.language_model.layers.47.self_attn.o_proj.weight",
508
+ "model.language_model.layers.47.self_attn.q_proj.weight",
509
+ "model.language_model.layers.47.self_attn.v_proj.weight"
510
+ ]
511
+ },
512
+ "modules_to_not_convert": [
513
+ ".condition_embedder",
514
+ ".txt_out",
515
+ "lm_head.weight",
516
+ ".vid_out",
517
+ ".emb_out",
518
+ ".emb_in",
519
+ "model.language_model.embed_tokens",
520
+ "lm_head",
521
+ ".img_out",
522
+ ".t_embedder",
523
+ ".norm_out",
524
+ ".txt_in",
525
+ "time_text_embed",
526
+ ".context_embedder",
527
+ ".y_embedder",
528
+ "patch_emb",
529
+ "wte",
530
+ ".final_layer",
531
+ ".img_in",
532
+ ".x_embedder",
533
+ "multi_modal_projector",
534
+ "patch_embed",
535
+ ".time_embed",
536
+ "model.language_model.norm",
537
+ ".vid_in",
538
+ ".proj_out",
539
+ "patch_embedding"
540
+ ],
541
+ "non_blocking": false,
542
+ "quant_conv": false,
543
+ "quant_method": "sdnq",
544
+ "quantization_device": null,
545
+ "quantized_matmul_dtype": "int8",
546
+ "return_device": null,
547
+ "sdnq_version": "0.1.4",
548
+ "svd_rank": 32,
549
+ "svd_steps": 8,
550
+ "use_dynamic_quantization": true,
551
+ "use_grad_ckpt": true,
552
+ "use_quantized_matmul": false,
553
+ "use_quantized_matmul_conv": false,
554
+ "use_static_quantization": true,
555
+ "use_stochastic_rounding": false,
556
+ "use_svd": false,
557
+ "weights_dtype": "int4"
558
+ }
tokenizer/added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144
3
+ }
tokenizer/chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
tokenizer/special_tokens_map.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "boi_token": "<start_of_image>",
3
+ "bos_token": {
4
+ "content": "<bos>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ "eoi_token": "<end_of_image>",
11
+ "eos_token": {
12
+ "content": "<eos>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false
17
+ },
18
+ "image_token": "<image_soft_token>",
19
+ "pad_token": {
20
+ "content": "<pad>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false
25
+ },
26
+ "unk_token": {
27
+ "content": "<unk>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
tokenizer/tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67ba3d0484274397fa3af834be403b419da0e1e35f7be6a2fcd2d9d67b700fc1
3
+ size 33384834
tokenizer/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
tokenizer/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff
 
transformer/config.json ADDED
@@ -0,0 +1,1436 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "LTX2VideoTransformer3DModel",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "Lightricks/LTX-2",
5
+ "activation_fn": "gelu-approximate",
6
+ "attention_bias": true,
7
+ "attention_head_dim": 128,
8
+ "attention_out_bias": true,
9
+ "audio_attention_head_dim": 64,
10
+ "audio_cross_attention_dim": 2048,
11
+ "audio_hop_length": 160,
12
+ "audio_in_channels": 128,
13
+ "audio_num_attention_heads": 32,
14
+ "audio_out_channels": 128,
15
+ "audio_patch_size": 1,
16
+ "audio_patch_size_t": 1,
17
+ "audio_pos_embed_max_pos": 20,
18
+ "audio_sampling_rate": 16000,
19
+ "audio_scale_factor": 4,
20
+ "base_height": 2048,
21
+ "base_width": 2048,
22
+ "caption_channels": 3840,
23
+ "causal_offset": 1,
24
+ "cross_attention_dim": 4096,
25
+ "cross_attn_timestep_scale_multiplier": 1000,
26
+ "in_channels": 128,
27
+ "norm_elementwise_affine": false,
28
+ "norm_eps": 1e-06,
29
+ "num_attention_heads": 32,
30
+ "num_layers": 48,
31
+ "out_channels": 128,
32
+ "patch_size": 1,
33
+ "patch_size_t": 1,
34
+ "pos_embed_max_pos": 20,
35
+ "qk_norm": "rms_norm_across_heads",
36
+ "quantization_config": {
37
+ "add_skip_keys": true,
38
+ "dequantize_fp32": false,
39
+ "dynamic_loss_threshold": 0.01,
40
+ "group_size": 0,
41
+ "is_integer": true,
42
+ "is_training": false,
43
+ "modules_dtype_dict": {
44
+ "float5_e3m1fn": [
45
+ "transformer_blocks.0.attn1.to_k.weight",
46
+ "transformer_blocks.0.attn1.to_v.weight",
47
+ "transformer_blocks.0.ff.net.2.weight",
48
+ "transformer_blocks.0.video_to_audio_attn.to_k.weight",
49
+ "transformer_blocks.0.video_to_audio_attn.to_v.weight",
50
+ "transformer_blocks.1.ff.net.0.proj.weight",
51
+ "transformer_blocks.3.ff.net.2.weight"
52
+ ],
53
+ "int5": [
54
+ "transformer_blocks.0.attn1.to_q.weight",
55
+ "transformer_blocks.0.audio_attn1.to_k.weight",
56
+ "transformer_blocks.0.audio_attn1.to_out.0.weight",
57
+ "transformer_blocks.0.audio_attn1.to_q.weight",
58
+ "transformer_blocks.0.audio_attn1.to_v.weight",
59
+ "transformer_blocks.0.audio_attn2.to_out.0.weight",
60
+ "transformer_blocks.0.audio_attn2.to_q.weight",
61
+ "transformer_blocks.0.audio_ff.net.0.proj.weight",
62
+ "transformer_blocks.0.audio_to_video_attn.to_k.weight",
63
+ "transformer_blocks.0.audio_to_video_attn.to_q.weight",
64
+ "transformer_blocks.0.audio_to_video_attn.to_v.weight",
65
+ "transformer_blocks.0.ff.net.0.proj.weight",
66
+ "transformer_blocks.0.video_to_audio_attn.to_q.weight",
67
+ "transformer_blocks.1.attn1.to_k.weight",
68
+ "transformer_blocks.1.attn1.to_q.weight",
69
+ "transformer_blocks.1.attn1.to_v.weight",
70
+ "transformer_blocks.1.audio_attn1.to_k.weight",
71
+ "transformer_blocks.1.audio_attn1.to_q.weight",
72
+ "transformer_blocks.1.audio_attn1.to_v.weight",
73
+ "transformer_blocks.1.audio_attn2.to_q.weight",
74
+ "transformer_blocks.1.audio_ff.net.0.proj.weight",
75
+ "transformer_blocks.1.audio_ff.net.2.weight",
76
+ "transformer_blocks.1.audio_to_video_attn.to_k.weight",
77
+ "transformer_blocks.1.audio_to_video_attn.to_q.weight",
78
+ "transformer_blocks.1.audio_to_video_attn.to_v.weight",
79
+ "transformer_blocks.1.ff.net.2.weight",
80
+ "transformer_blocks.1.video_to_audio_attn.to_k.weight",
81
+ "transformer_blocks.1.video_to_audio_attn.to_q.weight",
82
+ "transformer_blocks.1.video_to_audio_attn.to_v.weight",
83
+ "transformer_blocks.2.attn1.to_k.weight",
84
+ "transformer_blocks.2.attn1.to_q.weight",
85
+ "transformer_blocks.2.attn1.to_v.weight",
86
+ "transformer_blocks.2.audio_attn1.to_k.weight",
87
+ "transformer_blocks.2.audio_attn1.to_q.weight",
88
+ "transformer_blocks.2.audio_attn1.to_v.weight",
89
+ "transformer_blocks.2.audio_attn2.to_q.weight",
90
+ "transformer_blocks.2.audio_ff.net.2.weight",
91
+ "transformer_blocks.2.audio_to_video_attn.to_k.weight",
92
+ "transformer_blocks.2.audio_to_video_attn.to_q.weight",
93
+ "transformer_blocks.2.audio_to_video_attn.to_v.weight",
94
+ "transformer_blocks.2.ff.net.0.proj.weight",
95
+ "transformer_blocks.2.ff.net.2.weight",
96
+ "transformer_blocks.2.video_to_audio_attn.to_k.weight",
97
+ "transformer_blocks.2.video_to_audio_attn.to_q.weight",
98
+ "transformer_blocks.2.video_to_audio_attn.to_v.weight",
99
+ "transformer_blocks.3.attn1.to_k.weight",
100
+ "transformer_blocks.3.attn1.to_v.weight",
101
+ "transformer_blocks.3.audio_attn1.to_k.weight",
102
+ "transformer_blocks.3.audio_attn1.to_v.weight",
103
+ "transformer_blocks.3.audio_attn2.to_q.weight",
104
+ "transformer_blocks.3.audio_ff.net.2.weight",
105
+ "transformer_blocks.3.audio_to_video_attn.to_k.weight",
106
+ "transformer_blocks.3.audio_to_video_attn.to_q.weight",
107
+ "transformer_blocks.3.audio_to_video_attn.to_v.weight",
108
+ "transformer_blocks.3.ff.net.0.proj.weight",
109
+ "transformer_blocks.3.video_to_audio_attn.to_k.weight",
110
+ "transformer_blocks.3.video_to_audio_attn.to_v.weight",
111
+ "transformer_blocks.4.attn1.to_v.weight",
112
+ "transformer_blocks.4.attn2.to_q.weight",
113
+ "transformer_blocks.4.audio_attn1.to_k.weight",
114
+ "transformer_blocks.4.audio_attn1.to_v.weight",
115
+ "transformer_blocks.4.audio_attn2.to_q.weight",
116
+ "transformer_blocks.4.audio_ff.net.2.weight",
117
+ "transformer_blocks.4.audio_to_video_attn.to_k.weight",
118
+ "transformer_blocks.4.audio_to_video_attn.to_v.weight",
119
+ "transformer_blocks.4.ff.net.0.proj.weight",
120
+ "transformer_blocks.4.ff.net.2.weight",
121
+ "transformer_blocks.4.video_to_audio_attn.to_k.weight",
122
+ "transformer_blocks.4.video_to_audio_attn.to_v.weight",
123
+ "transformer_blocks.5.attn1.to_k.weight",
124
+ "transformer_blocks.5.attn1.to_v.weight",
125
+ "transformer_blocks.5.audio_attn2.to_q.weight",
126
+ "transformer_blocks.5.audio_to_video_attn.to_q.weight",
127
+ "transformer_blocks.5.video_to_audio_attn.to_k.weight",
128
+ "transformer_blocks.5.video_to_audio_attn.to_v.weight",
129
+ "transformer_blocks.10.attn2.to_k.weight",
130
+ "transformer_blocks.10.attn2.to_q.weight",
131
+ "transformer_blocks.10.audio_ff.net.2.weight",
132
+ "transformer_blocks.10.ff.net.2.weight",
133
+ "transformer_blocks.11.attn2.to_k.weight",
134
+ "transformer_blocks.11.audio_ff.net.2.weight",
135
+ "transformer_blocks.11.ff.net.2.weight",
136
+ "transformer_blocks.5.audio_ff.net.2.weight",
137
+ "transformer_blocks.5.ff.net.2.weight",
138
+ "transformer_blocks.6.audio_attn2.to_q.weight",
139
+ "transformer_blocks.6.audio_ff.net.2.weight",
140
+ "transformer_blocks.6.ff.net.2.weight",
141
+ "transformer_blocks.7.attn2.to_k.weight",
142
+ "transformer_blocks.7.attn2.to_q.weight",
143
+ "transformer_blocks.7.audio_ff.net.2.weight",
144
+ "transformer_blocks.7.ff.net.2.weight",
145
+ "transformer_blocks.8.audio_attn1.to_v.weight",
146
+ "transformer_blocks.8.audio_ff.net.2.weight",
147
+ "transformer_blocks.8.ff.net.2.weight",
148
+ "transformer_blocks.9.attn1.to_v.weight",
149
+ "transformer_blocks.9.audio_ff.net.2.weight",
150
+ "transformer_blocks.9.ff.net.2.weight",
151
+ "transformer_blocks.12.attn1.to_v.weight",
152
+ "transformer_blocks.12.attn2.to_k.weight",
153
+ "transformer_blocks.12.audio_ff.net.2.weight",
154
+ "transformer_blocks.12.ff.net.2.weight",
155
+ "transformer_blocks.13.audio_ff.net.2.weight",
156
+ "transformer_blocks.13.ff.net.2.weight",
157
+ "transformer_blocks.14.audio_ff.net.2.weight",
158
+ "transformer_blocks.14.ff.net.2.weight",
159
+ "transformer_blocks.15.audio_ff.net.2.weight",
160
+ "transformer_blocks.15.ff.net.2.weight",
161
+ "transformer_blocks.16.audio_ff.net.2.weight",
162
+ "transformer_blocks.16.ff.net.2.weight",
163
+ "transformer_blocks.17.audio_ff.net.2.weight",
164
+ "transformer_blocks.17.ff.net.2.weight",
165
+ "transformer_blocks.18.attn2.to_k.weight",
166
+ "transformer_blocks.18.ff.net.2.weight",
167
+ "transformer_blocks.19.attn2.to_k.weight",
168
+ "transformer_blocks.19.ff.net.2.weight",
169
+ "transformer_blocks.20.attn2.to_k.weight",
170
+ "transformer_blocks.20.ff.net.2.weight",
171
+ "transformer_blocks.21.attn2.to_k.weight",
172
+ "transformer_blocks.22.attn2.to_k.weight",
173
+ "transformer_blocks.23.attn2.to_k.weight",
174
+ "transformer_blocks.23.attn2.to_v.weight",
175
+ "transformer_blocks.24.attn2.to_k.weight",
176
+ "transformer_blocks.24.attn2.to_v.weight",
177
+ "transformer_blocks.25.attn2.to_k.weight",
178
+ "transformer_blocks.25.attn2.to_v.weight",
179
+ "transformer_blocks.25.audio_attn1.to_v.weight",
180
+ "transformer_blocks.25.audio_to_video_attn.to_v.weight",
181
+ "transformer_blocks.26.attn2.to_k.weight",
182
+ "transformer_blocks.26.attn2.to_v.weight",
183
+ "transformer_blocks.27.attn2.to_k.weight",
184
+ "transformer_blocks.27.attn2.to_v.weight",
185
+ "transformer_blocks.27.audio_attn1.to_v.weight",
186
+ "transformer_blocks.27.audio_to_video_attn.to_v.weight",
187
+ "transformer_blocks.28.attn2.to_k.weight",
188
+ "transformer_blocks.28.audio_attn1.to_v.weight",
189
+ "transformer_blocks.29.attn1.to_k.weight",
190
+ "transformer_blocks.29.attn2.to_k.weight",
191
+ "transformer_blocks.29.attn2.to_q.weight",
192
+ "transformer_blocks.29.attn2.to_v.weight",
193
+ "transformer_blocks.29.video_to_audio_attn.to_v.weight",
194
+ "transformer_blocks.30.attn1.to_k.weight",
195
+ "transformer_blocks.30.attn2.to_k.weight",
196
+ "transformer_blocks.30.attn2.to_v.weight",
197
+ "transformer_blocks.31.attn1.to_k.weight",
198
+ "transformer_blocks.31.attn2.to_k.weight",
199
+ "transformer_blocks.31.attn2.to_v.weight",
200
+ "transformer_blocks.32.attn2.to_k.weight",
201
+ "transformer_blocks.32.attn2.to_v.weight",
202
+ "transformer_blocks.32.video_to_audio_attn.to_v.weight",
203
+ "transformer_blocks.33.attn1.to_k.weight",
204
+ "transformer_blocks.33.attn2.to_k.weight",
205
+ "transformer_blocks.33.attn2.to_v.weight",
206
+ "transformer_blocks.34.attn2.to_k.weight",
207
+ "transformer_blocks.34.attn2.to_out.0.weight",
208
+ "transformer_blocks.34.attn2.to_v.weight",
209
+ "transformer_blocks.35.attn2.to_k.weight",
210
+ "transformer_blocks.35.attn2.to_out.0.weight",
211
+ "transformer_blocks.35.attn2.to_v.weight",
212
+ "transformer_blocks.36.attn2.to_k.weight",
213
+ "transformer_blocks.36.attn2.to_out.0.weight",
214
+ "transformer_blocks.36.attn2.to_v.weight",
215
+ "transformer_blocks.36.audio_attn1.to_v.weight",
216
+ "transformer_blocks.36.audio_to_video_attn.to_v.weight",
217
+ "transformer_blocks.37.attn1.to_k.weight",
218
+ "transformer_blocks.37.attn1.to_v.weight",
219
+ "transformer_blocks.37.attn2.to_k.weight",
220
+ "transformer_blocks.37.attn2.to_out.0.weight",
221
+ "transformer_blocks.37.attn2.to_v.weight",
222
+ "transformer_blocks.38.attn1.to_k.weight",
223
+ "transformer_blocks.38.attn1.to_v.weight",
224
+ "transformer_blocks.38.attn2.to_k.weight",
225
+ "transformer_blocks.38.attn2.to_out.0.weight",
226
+ "transformer_blocks.38.attn2.to_v.weight",
227
+ "transformer_blocks.38.audio_attn1.to_v.weight",
228
+ "transformer_blocks.38.audio_to_video_attn.to_v.weight",
229
+ "transformer_blocks.39.attn1.to_k.weight",
230
+ "transformer_blocks.39.attn1.to_v.weight",
231
+ "transformer_blocks.39.attn2.to_k.weight",
232
+ "transformer_blocks.39.attn2.to_out.0.weight",
233
+ "transformer_blocks.39.attn2.to_v.weight",
234
+ "transformer_blocks.40.attn1.to_k.weight",
235
+ "transformer_blocks.40.attn1.to_v.weight",
236
+ "transformer_blocks.40.attn2.to_k.weight",
237
+ "transformer_blocks.40.attn2.to_out.0.weight",
238
+ "transformer_blocks.40.attn2.to_v.weight",
239
+ "transformer_blocks.40.video_to_audio_attn.to_k.weight",
240
+ "transformer_blocks.41.attn1.to_k.weight",
241
+ "transformer_blocks.41.attn1.to_v.weight",
242
+ "transformer_blocks.41.attn2.to_k.weight",
243
+ "transformer_blocks.41.attn2.to_out.0.weight",
244
+ "transformer_blocks.41.attn2.to_v.weight",
245
+ "transformer_blocks.41.audio_attn1.to_v.weight",
246
+ "transformer_blocks.41.audio_to_video_attn.to_v.weight",
247
+ "transformer_blocks.41.video_to_audio_attn.to_k.weight",
248
+ "transformer_blocks.42.attn1.to_k.weight",
249
+ "transformer_blocks.42.attn1.to_v.weight",
250
+ "transformer_blocks.42.attn2.to_k.weight",
251
+ "transformer_blocks.42.attn2.to_out.0.weight",
252
+ "transformer_blocks.42.attn2.to_v.weight",
253
+ "transformer_blocks.42.audio_ff.net.2.weight",
254
+ "transformer_blocks.42.audio_to_video_attn.to_v.weight",
255
+ "transformer_blocks.42.video_to_audio_attn.to_k.weight",
256
+ "transformer_blocks.43.attn1.to_k.weight",
257
+ "transformer_blocks.43.attn2.to_k.weight",
258
+ "transformer_blocks.43.attn2.to_out.0.weight",
259
+ "transformer_blocks.43.attn2.to_v.weight",
260
+ "transformer_blocks.43.audio_ff.net.2.weight",
261
+ "transformer_blocks.43.video_to_audio_attn.to_k.weight",
262
+ "transformer_blocks.44.attn1.to_k.weight",
263
+ "transformer_blocks.44.attn2.to_k.weight",
264
+ "transformer_blocks.44.attn2.to_out.0.weight",
265
+ "transformer_blocks.44.attn2.to_v.weight",
266
+ "transformer_blocks.44.audio_ff.net.2.weight",
267
+ "transformer_blocks.44.video_to_audio_attn.to_k.weight",
268
+ "transformer_blocks.45.attn1.to_k.weight",
269
+ "transformer_blocks.45.attn1.to_v.weight",
270
+ "transformer_blocks.45.attn2.to_k.weight",
271
+ "transformer_blocks.45.attn2.to_out.0.weight",
272
+ "transformer_blocks.45.attn2.to_v.weight",
273
+ "transformer_blocks.45.audio_ff.net.2.weight",
274
+ "transformer_blocks.45.video_to_audio_attn.to_k.weight",
275
+ "transformer_blocks.46.attn1.to_k.weight",
276
+ "transformer_blocks.46.attn1.to_v.weight",
277
+ "transformer_blocks.46.attn2.to_k.weight",
278
+ "transformer_blocks.46.attn2.to_out.0.weight",
279
+ "transformer_blocks.46.attn2.to_v.weight",
280
+ "transformer_blocks.46.audio_attn1.to_k.weight",
281
+ "transformer_blocks.46.audio_attn1.to_v.weight",
282
+ "transformer_blocks.46.audio_ff.net.2.weight",
283
+ "transformer_blocks.46.audio_to_video_attn.to_k.weight",
284
+ "transformer_blocks.46.audio_to_video_attn.to_v.weight",
285
+ "transformer_blocks.46.video_to_audio_attn.to_k.weight",
286
+ "transformer_blocks.47.attn1.to_k.weight",
287
+ "transformer_blocks.47.attn1.to_q.weight",
288
+ "transformer_blocks.47.attn1.to_v.weight",
289
+ "transformer_blocks.47.attn2.to_out.0.weight",
290
+ "transformer_blocks.47.attn2.to_v.weight",
291
+ "transformer_blocks.47.audio_attn1.to_v.weight",
292
+ "transformer_blocks.47.audio_attn2.to_out.0.weight",
293
+ "transformer_blocks.47.audio_ff.net.2.weight",
294
+ "transformer_blocks.47.audio_to_video_attn.to_q.weight",
295
+ "transformer_blocks.47.audio_to_video_attn.to_v.weight",
296
+ "transformer_blocks.47.ff.net.2.weight",
297
+ "transformer_blocks.47.video_to_audio_attn.to_k.weight",
298
+ "transformer_blocks.47.video_to_audio_attn.to_v.weight"
299
+ ],
300
+ "uint4": [
301
+ "transformer_blocks.0.attn1.to_out.0.weight",
302
+ "transformer_blocks.0.attn2.to_k.weight",
303
+ "transformer_blocks.0.attn2.to_out.0.weight",
304
+ "transformer_blocks.0.attn2.to_q.weight",
305
+ "transformer_blocks.0.attn2.to_v.weight",
306
+ "transformer_blocks.0.audio_attn2.to_k.weight",
307
+ "transformer_blocks.0.audio_attn2.to_v.weight",
308
+ "transformer_blocks.0.audio_ff.net.2.weight",
309
+ "transformer_blocks.0.audio_to_video_attn.to_out.0.weight",
310
+ "transformer_blocks.0.video_to_audio_attn.to_out.0.weight",
311
+ "transformer_blocks.1.attn1.to_out.0.weight",
312
+ "transformer_blocks.1.attn2.to_k.weight",
313
+ "transformer_blocks.1.attn2.to_out.0.weight",
314
+ "transformer_blocks.1.attn2.to_q.weight",
315
+ "transformer_blocks.1.attn2.to_v.weight",
316
+ "transformer_blocks.1.audio_attn1.to_out.0.weight",
317
+ "transformer_blocks.1.audio_attn2.to_k.weight",
318
+ "transformer_blocks.1.audio_attn2.to_out.0.weight",
319
+ "transformer_blocks.1.audio_attn2.to_v.weight",
320
+ "transformer_blocks.1.audio_to_video_attn.to_out.0.weight",
321
+ "transformer_blocks.1.video_to_audio_attn.to_out.0.weight",
322
+ "transformer_blocks.2.attn1.to_out.0.weight",
323
+ "transformer_blocks.2.attn2.to_k.weight",
324
+ "transformer_blocks.2.attn2.to_out.0.weight",
325
+ "transformer_blocks.2.attn2.to_q.weight",
326
+ "transformer_blocks.2.attn2.to_v.weight",
327
+ "transformer_blocks.2.audio_attn1.to_out.0.weight",
328
+ "transformer_blocks.2.audio_attn2.to_k.weight",
329
+ "transformer_blocks.2.audio_attn2.to_out.0.weight",
330
+ "transformer_blocks.2.audio_attn2.to_v.weight",
331
+ "transformer_blocks.2.audio_ff.net.0.proj.weight",
332
+ "transformer_blocks.2.audio_to_video_attn.to_out.0.weight",
333
+ "transformer_blocks.2.video_to_audio_attn.to_out.0.weight",
334
+ "transformer_blocks.3.attn1.to_out.0.weight",
335
+ "transformer_blocks.3.attn1.to_q.weight",
336
+ "transformer_blocks.3.attn2.to_k.weight",
337
+ "transformer_blocks.3.attn2.to_out.0.weight",
338
+ "transformer_blocks.3.attn2.to_q.weight",
339
+ "transformer_blocks.3.attn2.to_v.weight",
340
+ "transformer_blocks.3.audio_attn1.to_out.0.weight",
341
+ "transformer_blocks.3.audio_attn1.to_q.weight",
342
+ "transformer_blocks.3.audio_attn2.to_k.weight",
343
+ "transformer_blocks.3.audio_attn2.to_out.0.weight",
344
+ "transformer_blocks.3.audio_attn2.to_v.weight",
345
+ "transformer_blocks.3.audio_ff.net.0.proj.weight",
346
+ "transformer_blocks.3.audio_to_video_attn.to_out.0.weight",
347
+ "transformer_blocks.3.video_to_audio_attn.to_out.0.weight",
348
+ "transformer_blocks.3.video_to_audio_attn.to_q.weight",
349
+ "transformer_blocks.4.attn1.to_k.weight",
350
+ "transformer_blocks.4.attn1.to_out.0.weight",
351
+ "transformer_blocks.4.attn1.to_q.weight",
352
+ "transformer_blocks.4.attn2.to_k.weight",
353
+ "transformer_blocks.4.attn2.to_out.0.weight",
354
+ "transformer_blocks.4.attn2.to_v.weight",
355
+ "transformer_blocks.4.audio_attn1.to_out.0.weight",
356
+ "transformer_blocks.4.audio_attn1.to_q.weight",
357
+ "transformer_blocks.4.audio_attn2.to_k.weight",
358
+ "transformer_blocks.4.audio_attn2.to_out.0.weight",
359
+ "transformer_blocks.4.audio_attn2.to_v.weight",
360
+ "transformer_blocks.4.audio_ff.net.0.proj.weight",
361
+ "transformer_blocks.4.audio_to_video_attn.to_out.0.weight",
362
+ "transformer_blocks.4.audio_to_video_attn.to_q.weight",
363
+ "transformer_blocks.4.video_to_audio_attn.to_out.0.weight",
364
+ "transformer_blocks.4.video_to_audio_attn.to_q.weight",
365
+ "transformer_blocks.5.attn1.to_out.0.weight",
366
+ "transformer_blocks.5.attn1.to_q.weight",
367
+ "transformer_blocks.5.attn2.to_k.weight",
368
+ "transformer_blocks.5.attn2.to_out.0.weight",
369
+ "transformer_blocks.5.attn2.to_q.weight",
370
+ "transformer_blocks.5.attn2.to_v.weight",
371
+ "transformer_blocks.5.audio_attn1.to_k.weight",
372
+ "transformer_blocks.5.audio_attn1.to_out.0.weight",
373
+ "transformer_blocks.5.audio_attn1.to_q.weight",
374
+ "transformer_blocks.5.audio_attn1.to_v.weight",
375
+ "transformer_blocks.5.audio_attn2.to_k.weight",
376
+ "transformer_blocks.5.audio_attn2.to_out.0.weight",
377
+ "transformer_blocks.5.audio_attn2.to_v.weight",
378
+ "transformer_blocks.5.audio_to_video_attn.to_k.weight",
379
+ "transformer_blocks.5.audio_to_video_attn.to_out.0.weight",
380
+ "transformer_blocks.5.audio_to_video_attn.to_v.weight",
381
+ "transformer_blocks.5.video_to_audio_attn.to_out.0.weight",
382
+ "transformer_blocks.5.video_to_audio_attn.to_q.weight",
383
+ "transformer_blocks.10.attn1.to_k.weight",
384
+ "transformer_blocks.10.attn1.to_out.0.weight",
385
+ "transformer_blocks.10.attn1.to_q.weight",
386
+ "transformer_blocks.10.attn1.to_v.weight",
387
+ "transformer_blocks.10.attn2.to_out.0.weight",
388
+ "transformer_blocks.10.attn2.to_v.weight",
389
+ "transformer_blocks.10.audio_attn1.to_k.weight",
390
+ "transformer_blocks.10.audio_attn1.to_out.0.weight",
391
+ "transformer_blocks.10.audio_attn1.to_q.weight",
392
+ "transformer_blocks.10.audio_attn1.to_v.weight",
393
+ "transformer_blocks.10.audio_attn2.to_k.weight",
394
+ "transformer_blocks.10.audio_attn2.to_out.0.weight",
395
+ "transformer_blocks.10.audio_attn2.to_q.weight",
396
+ "transformer_blocks.10.audio_attn2.to_v.weight",
397
+ "transformer_blocks.10.audio_ff.net.0.proj.weight",
398
+ "transformer_blocks.10.audio_to_video_attn.to_k.weight",
399
+ "transformer_blocks.10.audio_to_video_attn.to_out.0.weight",
400
+ "transformer_blocks.10.audio_to_video_attn.to_q.weight",
401
+ "transformer_blocks.10.audio_to_video_attn.to_v.weight",
402
+ "transformer_blocks.10.ff.net.0.proj.weight",
403
+ "transformer_blocks.10.video_to_audio_attn.to_k.weight",
404
+ "transformer_blocks.10.video_to_audio_attn.to_out.0.weight",
405
+ "transformer_blocks.10.video_to_audio_attn.to_q.weight",
406
+ "transformer_blocks.10.video_to_audio_attn.to_v.weight",
407
+ "transformer_blocks.11.attn1.to_k.weight",
408
+ "transformer_blocks.11.attn1.to_out.0.weight",
409
+ "transformer_blocks.11.attn1.to_q.weight",
410
+ "transformer_blocks.11.attn1.to_v.weight",
411
+ "transformer_blocks.11.attn2.to_out.0.weight",
412
+ "transformer_blocks.11.attn2.to_q.weight",
413
+ "transformer_blocks.11.attn2.to_v.weight",
414
+ "transformer_blocks.11.audio_attn1.to_k.weight",
415
+ "transformer_blocks.11.audio_attn1.to_out.0.weight",
416
+ "transformer_blocks.11.audio_attn1.to_q.weight",
417
+ "transformer_blocks.11.audio_attn1.to_v.weight",
418
+ "transformer_blocks.11.audio_attn2.to_k.weight",
419
+ "transformer_blocks.11.audio_attn2.to_out.0.weight",
420
+ "transformer_blocks.11.audio_attn2.to_q.weight",
421
+ "transformer_blocks.11.audio_attn2.to_v.weight",
422
+ "transformer_blocks.11.audio_ff.net.0.proj.weight",
423
+ "transformer_blocks.11.audio_to_video_attn.to_k.weight",
424
+ "transformer_blocks.11.audio_to_video_attn.to_out.0.weight",
425
+ "transformer_blocks.11.audio_to_video_attn.to_q.weight",
426
+ "transformer_blocks.11.audio_to_video_attn.to_v.weight",
427
+ "transformer_blocks.11.ff.net.0.proj.weight",
428
+ "transformer_blocks.11.video_to_audio_attn.to_k.weight",
429
+ "transformer_blocks.11.video_to_audio_attn.to_out.0.weight",
430
+ "transformer_blocks.11.video_to_audio_attn.to_q.weight",
431
+ "transformer_blocks.11.video_to_audio_attn.to_v.weight",
432
+ "transformer_blocks.5.audio_ff.net.0.proj.weight",
433
+ "transformer_blocks.5.ff.net.0.proj.weight",
434
+ "transformer_blocks.6.attn1.to_k.weight",
435
+ "transformer_blocks.6.attn1.to_out.0.weight",
436
+ "transformer_blocks.6.attn1.to_q.weight",
437
+ "transformer_blocks.6.attn1.to_v.weight",
438
+ "transformer_blocks.6.attn2.to_k.weight",
439
+ "transformer_blocks.6.attn2.to_out.0.weight",
440
+ "transformer_blocks.6.attn2.to_q.weight",
441
+ "transformer_blocks.6.attn2.to_v.weight",
442
+ "transformer_blocks.6.audio_attn1.to_k.weight",
443
+ "transformer_blocks.6.audio_attn1.to_out.0.weight",
444
+ "transformer_blocks.6.audio_attn1.to_q.weight",
445
+ "transformer_blocks.6.audio_attn1.to_v.weight",
446
+ "transformer_blocks.6.audio_attn2.to_k.weight",
447
+ "transformer_blocks.6.audio_attn2.to_out.0.weight",
448
+ "transformer_blocks.6.audio_attn2.to_v.weight",
449
+ "transformer_blocks.6.audio_ff.net.0.proj.weight",
450
+ "transformer_blocks.6.audio_to_video_attn.to_k.weight",
451
+ "transformer_blocks.6.audio_to_video_attn.to_out.0.weight",
452
+ "transformer_blocks.6.audio_to_video_attn.to_q.weight",
453
+ "transformer_blocks.6.audio_to_video_attn.to_v.weight",
454
+ "transformer_blocks.6.ff.net.0.proj.weight",
455
+ "transformer_blocks.6.video_to_audio_attn.to_k.weight",
456
+ "transformer_blocks.6.video_to_audio_attn.to_out.0.weight",
457
+ "transformer_blocks.6.video_to_audio_attn.to_q.weight",
458
+ "transformer_blocks.6.video_to_audio_attn.to_v.weight",
459
+ "transformer_blocks.7.attn1.to_k.weight",
460
+ "transformer_blocks.7.attn1.to_out.0.weight",
461
+ "transformer_blocks.7.attn1.to_q.weight",
462
+ "transformer_blocks.7.attn1.to_v.weight",
463
+ "transformer_blocks.7.attn2.to_out.0.weight",
464
+ "transformer_blocks.7.attn2.to_v.weight",
465
+ "transformer_blocks.7.audio_attn1.to_k.weight",
466
+ "transformer_blocks.7.audio_attn1.to_out.0.weight",
467
+ "transformer_blocks.7.audio_attn1.to_q.weight",
468
+ "transformer_blocks.7.audio_attn1.to_v.weight",
469
+ "transformer_blocks.7.audio_attn2.to_k.weight",
470
+ "transformer_blocks.7.audio_attn2.to_out.0.weight",
471
+ "transformer_blocks.7.audio_attn2.to_q.weight",
472
+ "transformer_blocks.7.audio_attn2.to_v.weight",
473
+ "transformer_blocks.7.audio_ff.net.0.proj.weight",
474
+ "transformer_blocks.7.audio_to_video_attn.to_k.weight",
475
+ "transformer_blocks.7.audio_to_video_attn.to_out.0.weight",
476
+ "transformer_blocks.7.audio_to_video_attn.to_q.weight",
477
+ "transformer_blocks.7.audio_to_video_attn.to_v.weight",
478
+ "transformer_blocks.7.ff.net.0.proj.weight",
479
+ "transformer_blocks.7.video_to_audio_attn.to_k.weight",
480
+ "transformer_blocks.7.video_to_audio_attn.to_out.0.weight",
481
+ "transformer_blocks.7.video_to_audio_attn.to_q.weight",
482
+ "transformer_blocks.7.video_to_audio_attn.to_v.weight",
483
+ "transformer_blocks.8.attn1.to_k.weight",
484
+ "transformer_blocks.8.attn1.to_out.0.weight",
485
+ "transformer_blocks.8.attn1.to_q.weight",
486
+ "transformer_blocks.8.attn1.to_v.weight",
487
+ "transformer_blocks.8.attn2.to_k.weight",
488
+ "transformer_blocks.8.attn2.to_out.0.weight",
489
+ "transformer_blocks.8.attn2.to_q.weight",
490
+ "transformer_blocks.8.attn2.to_v.weight",
491
+ "transformer_blocks.8.audio_attn1.to_k.weight",
492
+ "transformer_blocks.8.audio_attn1.to_out.0.weight",
493
+ "transformer_blocks.8.audio_attn1.to_q.weight",
494
+ "transformer_blocks.8.audio_attn2.to_k.weight",
495
+ "transformer_blocks.8.audio_attn2.to_out.0.weight",
496
+ "transformer_blocks.8.audio_attn2.to_q.weight",
497
+ "transformer_blocks.8.audio_attn2.to_v.weight",
498
+ "transformer_blocks.8.audio_ff.net.0.proj.weight",
499
+ "transformer_blocks.8.audio_to_video_attn.to_k.weight",
500
+ "transformer_blocks.8.audio_to_video_attn.to_out.0.weight",
501
+ "transformer_blocks.8.audio_to_video_attn.to_q.weight",
502
+ "transformer_blocks.8.audio_to_video_attn.to_v.weight",
503
+ "transformer_blocks.8.ff.net.0.proj.weight",
504
+ "transformer_blocks.8.video_to_audio_attn.to_k.weight",
505
+ "transformer_blocks.8.video_to_audio_attn.to_out.0.weight",
506
+ "transformer_blocks.8.video_to_audio_attn.to_q.weight",
507
+ "transformer_blocks.8.video_to_audio_attn.to_v.weight",
508
+ "transformer_blocks.9.attn1.to_k.weight",
509
+ "transformer_blocks.9.attn1.to_out.0.weight",
510
+ "transformer_blocks.9.attn1.to_q.weight",
511
+ "transformer_blocks.9.attn2.to_k.weight",
512
+ "transformer_blocks.9.attn2.to_out.0.weight",
513
+ "transformer_blocks.9.attn2.to_q.weight",
514
+ "transformer_blocks.9.attn2.to_v.weight",
515
+ "transformer_blocks.9.audio_attn1.to_k.weight",
516
+ "transformer_blocks.9.audio_attn1.to_out.0.weight",
517
+ "transformer_blocks.9.audio_attn1.to_q.weight",
518
+ "transformer_blocks.9.audio_attn1.to_v.weight",
519
+ "transformer_blocks.9.audio_attn2.to_k.weight",
520
+ "transformer_blocks.9.audio_attn2.to_out.0.weight",
521
+ "transformer_blocks.9.audio_attn2.to_q.weight",
522
+ "transformer_blocks.9.audio_attn2.to_v.weight",
523
+ "transformer_blocks.9.audio_ff.net.0.proj.weight",
524
+ "transformer_blocks.9.audio_to_video_attn.to_k.weight",
525
+ "transformer_blocks.9.audio_to_video_attn.to_out.0.weight",
526
+ "transformer_blocks.9.audio_to_video_attn.to_q.weight",
527
+ "transformer_blocks.9.audio_to_video_attn.to_v.weight",
528
+ "transformer_blocks.9.ff.net.0.proj.weight",
529
+ "transformer_blocks.9.video_to_audio_attn.to_k.weight",
530
+ "transformer_blocks.9.video_to_audio_attn.to_out.0.weight",
531
+ "transformer_blocks.9.video_to_audio_attn.to_q.weight",
532
+ "transformer_blocks.9.video_to_audio_attn.to_v.weight",
533
+ "transformer_blocks.12.attn1.to_k.weight",
534
+ "transformer_blocks.12.attn1.to_out.0.weight",
535
+ "transformer_blocks.12.attn1.to_q.weight",
536
+ "transformer_blocks.12.attn2.to_out.0.weight",
537
+ "transformer_blocks.12.attn2.to_q.weight",
538
+ "transformer_blocks.12.attn2.to_v.weight",
539
+ "transformer_blocks.12.audio_attn1.to_k.weight",
540
+ "transformer_blocks.12.audio_attn1.to_out.0.weight",
541
+ "transformer_blocks.12.audio_attn1.to_q.weight",
542
+ "transformer_blocks.12.audio_attn1.to_v.weight",
543
+ "transformer_blocks.12.audio_attn2.to_k.weight",
544
+ "transformer_blocks.12.audio_attn2.to_out.0.weight",
545
+ "transformer_blocks.12.audio_attn2.to_q.weight",
546
+ "transformer_blocks.12.audio_attn2.to_v.weight",
547
+ "transformer_blocks.12.audio_ff.net.0.proj.weight",
548
+ "transformer_blocks.12.audio_to_video_attn.to_k.weight",
549
+ "transformer_blocks.12.audio_to_video_attn.to_out.0.weight",
550
+ "transformer_blocks.12.audio_to_video_attn.to_q.weight",
551
+ "transformer_blocks.12.audio_to_video_attn.to_v.weight",
552
+ "transformer_blocks.12.ff.net.0.proj.weight",
553
+ "transformer_blocks.12.video_to_audio_attn.to_k.weight",
554
+ "transformer_blocks.12.video_to_audio_attn.to_out.0.weight",
555
+ "transformer_blocks.12.video_to_audio_attn.to_q.weight",
556
+ "transformer_blocks.12.video_to_audio_attn.to_v.weight",
557
+ "transformer_blocks.13.attn1.to_k.weight",
558
+ "transformer_blocks.13.attn1.to_out.0.weight",
559
+ "transformer_blocks.13.attn1.to_q.weight",
560
+ "transformer_blocks.13.attn1.to_v.weight",
561
+ "transformer_blocks.13.attn2.to_k.weight",
562
+ "transformer_blocks.13.attn2.to_out.0.weight",
563
+ "transformer_blocks.13.attn2.to_q.weight",
564
+ "transformer_blocks.13.attn2.to_v.weight",
565
+ "transformer_blocks.13.audio_attn1.to_k.weight",
566
+ "transformer_blocks.13.audio_attn1.to_out.0.weight",
567
+ "transformer_blocks.13.audio_attn1.to_q.weight",
568
+ "transformer_blocks.13.audio_attn1.to_v.weight",
569
+ "transformer_blocks.13.audio_attn2.to_k.weight",
570
+ "transformer_blocks.13.audio_attn2.to_out.0.weight",
571
+ "transformer_blocks.13.audio_attn2.to_q.weight",
572
+ "transformer_blocks.13.audio_attn2.to_v.weight",
573
+ "transformer_blocks.13.audio_ff.net.0.proj.weight",
574
+ "transformer_blocks.13.audio_to_video_attn.to_k.weight",
575
+ "transformer_blocks.13.audio_to_video_attn.to_out.0.weight",
576
+ "transformer_blocks.13.audio_to_video_attn.to_q.weight",
577
+ "transformer_blocks.13.audio_to_video_attn.to_v.weight",
578
+ "transformer_blocks.13.ff.net.0.proj.weight",
579
+ "transformer_blocks.13.video_to_audio_attn.to_k.weight",
580
+ "transformer_blocks.13.video_to_audio_attn.to_out.0.weight",
581
+ "transformer_blocks.13.video_to_audio_attn.to_q.weight",
582
+ "transformer_blocks.13.video_to_audio_attn.to_v.weight",
583
+ "transformer_blocks.14.attn1.to_k.weight",
584
+ "transformer_blocks.14.attn1.to_out.0.weight",
585
+ "transformer_blocks.14.attn1.to_q.weight",
586
+ "transformer_blocks.14.attn1.to_v.weight",
587
+ "transformer_blocks.14.attn2.to_k.weight",
588
+ "transformer_blocks.14.attn2.to_out.0.weight",
589
+ "transformer_blocks.14.attn2.to_q.weight",
590
+ "transformer_blocks.14.attn2.to_v.weight",
591
+ "transformer_blocks.14.audio_attn1.to_k.weight",
592
+ "transformer_blocks.14.audio_attn1.to_out.0.weight",
593
+ "transformer_blocks.14.audio_attn1.to_q.weight",
594
+ "transformer_blocks.14.audio_attn1.to_v.weight",
595
+ "transformer_blocks.14.audio_attn2.to_k.weight",
596
+ "transformer_blocks.14.audio_attn2.to_out.0.weight",
597
+ "transformer_blocks.14.audio_attn2.to_q.weight",
598
+ "transformer_blocks.14.audio_attn2.to_v.weight",
599
+ "transformer_blocks.14.audio_ff.net.0.proj.weight",
600
+ "transformer_blocks.14.audio_to_video_attn.to_k.weight",
601
+ "transformer_blocks.14.audio_to_video_attn.to_out.0.weight",
602
+ "transformer_blocks.14.audio_to_video_attn.to_q.weight",
603
+ "transformer_blocks.14.audio_to_video_attn.to_v.weight",
604
+ "transformer_blocks.14.ff.net.0.proj.weight",
605
+ "transformer_blocks.14.video_to_audio_attn.to_k.weight",
606
+ "transformer_blocks.14.video_to_audio_attn.to_out.0.weight",
607
+ "transformer_blocks.14.video_to_audio_attn.to_q.weight",
608
+ "transformer_blocks.14.video_to_audio_attn.to_v.weight",
609
+ "transformer_blocks.15.attn1.to_k.weight",
610
+ "transformer_blocks.15.attn1.to_out.0.weight",
611
+ "transformer_blocks.15.attn1.to_q.weight",
612
+ "transformer_blocks.15.attn1.to_v.weight",
613
+ "transformer_blocks.15.attn2.to_k.weight",
614
+ "transformer_blocks.15.attn2.to_out.0.weight",
615
+ "transformer_blocks.15.attn2.to_q.weight",
616
+ "transformer_blocks.15.attn2.to_v.weight",
617
+ "transformer_blocks.15.audio_attn1.to_k.weight",
618
+ "transformer_blocks.15.audio_attn1.to_out.0.weight",
619
+ "transformer_blocks.15.audio_attn1.to_q.weight",
620
+ "transformer_blocks.15.audio_attn1.to_v.weight",
621
+ "transformer_blocks.15.audio_attn2.to_k.weight",
622
+ "transformer_blocks.15.audio_attn2.to_out.0.weight",
623
+ "transformer_blocks.15.audio_attn2.to_q.weight",
624
+ "transformer_blocks.15.audio_attn2.to_v.weight",
625
+ "transformer_blocks.15.audio_ff.net.0.proj.weight",
626
+ "transformer_blocks.15.audio_to_video_attn.to_k.weight",
627
+ "transformer_blocks.15.audio_to_video_attn.to_out.0.weight",
628
+ "transformer_blocks.15.audio_to_video_attn.to_q.weight",
629
+ "transformer_blocks.15.audio_to_video_attn.to_v.weight",
630
+ "transformer_blocks.15.ff.net.0.proj.weight",
631
+ "transformer_blocks.15.video_to_audio_attn.to_k.weight",
632
+ "transformer_blocks.15.video_to_audio_attn.to_out.0.weight",
633
+ "transformer_blocks.15.video_to_audio_attn.to_q.weight",
634
+ "transformer_blocks.15.video_to_audio_attn.to_v.weight",
635
+ "transformer_blocks.16.attn1.to_k.weight",
636
+ "transformer_blocks.16.attn1.to_out.0.weight",
637
+ "transformer_blocks.16.attn1.to_q.weight",
638
+ "transformer_blocks.16.attn1.to_v.weight",
639
+ "transformer_blocks.16.attn2.to_k.weight",
640
+ "transformer_blocks.16.attn2.to_out.0.weight",
641
+ "transformer_blocks.16.attn2.to_q.weight",
642
+ "transformer_blocks.16.attn2.to_v.weight",
643
+ "transformer_blocks.16.audio_attn1.to_k.weight",
644
+ "transformer_blocks.16.audio_attn1.to_out.0.weight",
645
+ "transformer_blocks.16.audio_attn1.to_q.weight",
646
+ "transformer_blocks.16.audio_attn1.to_v.weight",
647
+ "transformer_blocks.16.audio_attn2.to_k.weight",
648
+ "transformer_blocks.16.audio_attn2.to_out.0.weight",
649
+ "transformer_blocks.16.audio_attn2.to_q.weight",
650
+ "transformer_blocks.16.audio_attn2.to_v.weight",
651
+ "transformer_blocks.16.audio_ff.net.0.proj.weight",
652
+ "transformer_blocks.16.audio_to_video_attn.to_k.weight",
653
+ "transformer_blocks.16.audio_to_video_attn.to_out.0.weight",
654
+ "transformer_blocks.16.audio_to_video_attn.to_q.weight",
655
+ "transformer_blocks.16.audio_to_video_attn.to_v.weight",
656
+ "transformer_blocks.16.ff.net.0.proj.weight",
657
+ "transformer_blocks.16.video_to_audio_attn.to_k.weight",
658
+ "transformer_blocks.16.video_to_audio_attn.to_out.0.weight",
659
+ "transformer_blocks.16.video_to_audio_attn.to_q.weight",
660
+ "transformer_blocks.16.video_to_audio_attn.to_v.weight",
661
+ "transformer_blocks.17.attn1.to_k.weight",
662
+ "transformer_blocks.17.attn1.to_out.0.weight",
663
+ "transformer_blocks.17.attn1.to_q.weight",
664
+ "transformer_blocks.17.attn1.to_v.weight",
665
+ "transformer_blocks.17.attn2.to_k.weight",
666
+ "transformer_blocks.17.attn2.to_out.0.weight",
667
+ "transformer_blocks.17.attn2.to_q.weight",
668
+ "transformer_blocks.17.attn2.to_v.weight",
669
+ "transformer_blocks.17.audio_attn1.to_k.weight",
670
+ "transformer_blocks.17.audio_attn1.to_out.0.weight",
671
+ "transformer_blocks.17.audio_attn1.to_q.weight",
672
+ "transformer_blocks.17.audio_attn1.to_v.weight",
673
+ "transformer_blocks.17.audio_attn2.to_k.weight",
674
+ "transformer_blocks.17.audio_attn2.to_out.0.weight",
675
+ "transformer_blocks.17.audio_attn2.to_q.weight",
676
+ "transformer_blocks.17.audio_attn2.to_v.weight",
677
+ "transformer_blocks.17.audio_ff.net.0.proj.weight",
678
+ "transformer_blocks.17.audio_to_video_attn.to_k.weight",
679
+ "transformer_blocks.17.audio_to_video_attn.to_out.0.weight",
680
+ "transformer_blocks.17.audio_to_video_attn.to_q.weight",
681
+ "transformer_blocks.17.audio_to_video_attn.to_v.weight",
682
+ "transformer_blocks.17.ff.net.0.proj.weight",
683
+ "transformer_blocks.17.video_to_audio_attn.to_k.weight",
684
+ "transformer_blocks.17.video_to_audio_attn.to_out.0.weight",
685
+ "transformer_blocks.17.video_to_audio_attn.to_q.weight",
686
+ "transformer_blocks.17.video_to_audio_attn.to_v.weight",
687
+ "transformer_blocks.18.attn1.to_k.weight",
688
+ "transformer_blocks.18.attn1.to_out.0.weight",
689
+ "transformer_blocks.18.attn1.to_q.weight",
690
+ "transformer_blocks.18.attn1.to_v.weight",
691
+ "transformer_blocks.18.attn2.to_out.0.weight",
692
+ "transformer_blocks.18.attn2.to_q.weight",
693
+ "transformer_blocks.18.attn2.to_v.weight",
694
+ "transformer_blocks.18.audio_attn1.to_k.weight",
695
+ "transformer_blocks.18.audio_attn1.to_out.0.weight",
696
+ "transformer_blocks.18.audio_attn1.to_q.weight",
697
+ "transformer_blocks.18.audio_attn1.to_v.weight",
698
+ "transformer_blocks.18.audio_attn2.to_k.weight",
699
+ "transformer_blocks.18.audio_attn2.to_out.0.weight",
700
+ "transformer_blocks.18.audio_attn2.to_q.weight",
701
+ "transformer_blocks.18.audio_attn2.to_v.weight",
702
+ "transformer_blocks.18.audio_to_video_attn.to_k.weight",
703
+ "transformer_blocks.18.audio_to_video_attn.to_q.weight",
704
+ "transformer_blocks.18.audio_ff.net.0.proj.weight",
705
+ "transformer_blocks.18.audio_ff.net.2.weight",
706
+ "transformer_blocks.18.audio_to_video_attn.to_out.0.weight",
707
+ "transformer_blocks.18.audio_to_video_attn.to_v.weight",
708
+ "transformer_blocks.18.ff.net.0.proj.weight",
709
+ "transformer_blocks.18.video_to_audio_attn.to_k.weight",
710
+ "transformer_blocks.18.video_to_audio_attn.to_out.0.weight",
711
+ "transformer_blocks.18.video_to_audio_attn.to_q.weight",
712
+ "transformer_blocks.18.video_to_audio_attn.to_v.weight",
713
+ "transformer_blocks.19.attn1.to_k.weight",
714
+ "transformer_blocks.19.attn1.to_out.0.weight",
715
+ "transformer_blocks.19.attn1.to_q.weight",
716
+ "transformer_blocks.19.attn1.to_v.weight",
717
+ "transformer_blocks.19.attn2.to_out.0.weight",
718
+ "transformer_blocks.19.attn2.to_q.weight",
719
+ "transformer_blocks.19.attn2.to_v.weight",
720
+ "transformer_blocks.19.audio_attn1.to_k.weight",
721
+ "transformer_blocks.19.audio_attn1.to_out.0.weight",
722
+ "transformer_blocks.19.audio_attn1.to_q.weight",
723
+ "transformer_blocks.19.audio_attn1.to_v.weight",
724
+ "transformer_blocks.19.audio_attn2.to_k.weight",
725
+ "transformer_blocks.19.audio_attn2.to_out.0.weight",
726
+ "transformer_blocks.19.audio_attn2.to_q.weight",
727
+ "transformer_blocks.19.audio_attn2.to_v.weight",
728
+ "transformer_blocks.19.audio_ff.net.0.proj.weight",
729
+ "transformer_blocks.19.audio_ff.net.2.weight",
730
+ "transformer_blocks.19.audio_to_video_attn.to_k.weight",
731
+ "transformer_blocks.19.audio_to_video_attn.to_out.0.weight",
732
+ "transformer_blocks.19.audio_to_video_attn.to_q.weight",
733
+ "transformer_blocks.19.audio_to_video_attn.to_v.weight",
734
+ "transformer_blocks.19.ff.net.0.proj.weight",
735
+ "transformer_blocks.19.video_to_audio_attn.to_k.weight",
736
+ "transformer_blocks.19.video_to_audio_attn.to_out.0.weight",
737
+ "transformer_blocks.19.video_to_audio_attn.to_q.weight",
738
+ "transformer_blocks.19.video_to_audio_attn.to_v.weight",
739
+ "transformer_blocks.20.attn1.to_k.weight",
740
+ "transformer_blocks.20.attn1.to_out.0.weight",
741
+ "transformer_blocks.20.attn1.to_q.weight",
742
+ "transformer_blocks.20.attn1.to_v.weight",
743
+ "transformer_blocks.20.attn2.to_out.0.weight",
744
+ "transformer_blocks.20.attn2.to_q.weight",
745
+ "transformer_blocks.20.attn2.to_v.weight",
746
+ "transformer_blocks.20.audio_attn1.to_k.weight",
747
+ "transformer_blocks.20.audio_attn1.to_out.0.weight",
748
+ "transformer_blocks.20.audio_attn1.to_q.weight",
749
+ "transformer_blocks.20.audio_attn1.to_v.weight",
750
+ "transformer_blocks.20.audio_attn2.to_k.weight",
751
+ "transformer_blocks.20.audio_attn2.to_out.0.weight",
752
+ "transformer_blocks.20.audio_attn2.to_q.weight",
753
+ "transformer_blocks.20.audio_attn2.to_v.weight",
754
+ "transformer_blocks.20.audio_ff.net.0.proj.weight",
755
+ "transformer_blocks.20.audio_ff.net.2.weight",
756
+ "transformer_blocks.20.audio_to_video_attn.to_k.weight",
757
+ "transformer_blocks.20.audio_to_video_attn.to_out.0.weight",
758
+ "transformer_blocks.20.audio_to_video_attn.to_q.weight",
759
+ "transformer_blocks.20.audio_to_video_attn.to_v.weight",
760
+ "transformer_blocks.20.ff.net.0.proj.weight",
761
+ "transformer_blocks.20.video_to_audio_attn.to_k.weight",
762
+ "transformer_blocks.20.video_to_audio_attn.to_out.0.weight",
763
+ "transformer_blocks.20.video_to_audio_attn.to_q.weight",
764
+ "transformer_blocks.20.video_to_audio_attn.to_v.weight",
765
+ "transformer_blocks.21.attn1.to_k.weight",
766
+ "transformer_blocks.21.attn1.to_out.0.weight",
767
+ "transformer_blocks.21.attn1.to_q.weight",
768
+ "transformer_blocks.21.attn1.to_v.weight",
769
+ "transformer_blocks.21.attn2.to_out.0.weight",
770
+ "transformer_blocks.21.attn2.to_q.weight",
771
+ "transformer_blocks.21.attn2.to_v.weight",
772
+ "transformer_blocks.21.audio_attn1.to_k.weight",
773
+ "transformer_blocks.21.audio_attn1.to_out.0.weight",
774
+ "transformer_blocks.21.audio_attn1.to_q.weight",
775
+ "transformer_blocks.21.audio_attn1.to_v.weight",
776
+ "transformer_blocks.21.audio_attn2.to_k.weight",
777
+ "transformer_blocks.21.audio_attn2.to_out.0.weight",
778
+ "transformer_blocks.21.audio_attn2.to_q.weight",
779
+ "transformer_blocks.21.audio_attn2.to_v.weight",
780
+ "transformer_blocks.21.audio_ff.net.0.proj.weight",
781
+ "transformer_blocks.21.audio_ff.net.2.weight",
782
+ "transformer_blocks.21.audio_to_video_attn.to_k.weight",
783
+ "transformer_blocks.21.audio_to_video_attn.to_out.0.weight",
784
+ "transformer_blocks.21.audio_to_video_attn.to_q.weight",
785
+ "transformer_blocks.21.audio_to_video_attn.to_v.weight",
786
+ "transformer_blocks.21.ff.net.0.proj.weight",
787
+ "transformer_blocks.21.ff.net.2.weight",
788
+ "transformer_blocks.21.video_to_audio_attn.to_k.weight",
789
+ "transformer_blocks.21.video_to_audio_attn.to_out.0.weight",
790
+ "transformer_blocks.21.video_to_audio_attn.to_q.weight",
791
+ "transformer_blocks.21.video_to_audio_attn.to_v.weight",
792
+ "transformer_blocks.22.attn1.to_k.weight",
793
+ "transformer_blocks.22.attn1.to_out.0.weight",
794
+ "transformer_blocks.22.attn1.to_q.weight",
795
+ "transformer_blocks.22.attn1.to_v.weight",
796
+ "transformer_blocks.22.attn2.to_out.0.weight",
797
+ "transformer_blocks.22.attn2.to_q.weight",
798
+ "transformer_blocks.22.attn2.to_v.weight",
799
+ "transformer_blocks.22.audio_attn1.to_k.weight",
800
+ "transformer_blocks.22.audio_attn1.to_out.0.weight",
801
+ "transformer_blocks.22.audio_attn1.to_q.weight",
802
+ "transformer_blocks.22.audio_attn1.to_v.weight",
803
+ "transformer_blocks.22.audio_attn2.to_k.weight",
804
+ "transformer_blocks.22.audio_attn2.to_out.0.weight",
805
+ "transformer_blocks.22.audio_attn2.to_q.weight",
806
+ "transformer_blocks.22.audio_attn2.to_v.weight",
807
+ "transformer_blocks.22.audio_ff.net.0.proj.weight",
808
+ "transformer_blocks.22.audio_ff.net.2.weight",
809
+ "transformer_blocks.22.audio_to_video_attn.to_k.weight",
810
+ "transformer_blocks.22.audio_to_video_attn.to_out.0.weight",
811
+ "transformer_blocks.22.audio_to_video_attn.to_q.weight",
812
+ "transformer_blocks.22.audio_to_video_attn.to_v.weight",
813
+ "transformer_blocks.22.ff.net.0.proj.weight",
814
+ "transformer_blocks.22.ff.net.2.weight",
815
+ "transformer_blocks.22.video_to_audio_attn.to_k.weight",
816
+ "transformer_blocks.22.video_to_audio_attn.to_out.0.weight",
817
+ "transformer_blocks.22.video_to_audio_attn.to_q.weight",
818
+ "transformer_blocks.22.video_to_audio_attn.to_v.weight",
819
+ "transformer_blocks.23.attn1.to_k.weight",
820
+ "transformer_blocks.23.attn1.to_out.0.weight",
821
+ "transformer_blocks.23.attn1.to_q.weight",
822
+ "transformer_blocks.23.attn1.to_v.weight",
823
+ "transformer_blocks.23.attn2.to_out.0.weight",
824
+ "transformer_blocks.23.attn2.to_q.weight",
825
+ "transformer_blocks.23.audio_attn1.to_k.weight",
826
+ "transformer_blocks.23.audio_attn1.to_out.0.weight",
827
+ "transformer_blocks.23.audio_attn1.to_q.weight",
828
+ "transformer_blocks.23.audio_attn1.to_v.weight",
829
+ "transformer_blocks.23.audio_attn2.to_k.weight",
830
+ "transformer_blocks.23.audio_attn2.to_out.0.weight",
831
+ "transformer_blocks.23.audio_attn2.to_q.weight",
832
+ "transformer_blocks.23.audio_attn2.to_v.weight",
833
+ "transformer_blocks.23.audio_ff.net.0.proj.weight",
834
+ "transformer_blocks.23.audio_ff.net.2.weight",
835
+ "transformer_blocks.23.audio_to_video_attn.to_k.weight",
836
+ "transformer_blocks.23.audio_to_video_attn.to_out.0.weight",
837
+ "transformer_blocks.23.audio_to_video_attn.to_q.weight",
838
+ "transformer_blocks.23.audio_to_video_attn.to_v.weight",
839
+ "transformer_blocks.23.ff.net.0.proj.weight",
840
+ "transformer_blocks.23.ff.net.2.weight",
841
+ "transformer_blocks.23.video_to_audio_attn.to_k.weight",
842
+ "transformer_blocks.23.video_to_audio_attn.to_out.0.weight",
843
+ "transformer_blocks.23.video_to_audio_attn.to_q.weight",
844
+ "transformer_blocks.23.video_to_audio_attn.to_v.weight",
845
+ "transformer_blocks.24.attn1.to_k.weight",
846
+ "transformer_blocks.24.attn1.to_out.0.weight",
847
+ "transformer_blocks.24.attn1.to_q.weight",
848
+ "transformer_blocks.24.attn1.to_v.weight",
849
+ "transformer_blocks.24.attn2.to_out.0.weight",
850
+ "transformer_blocks.24.attn2.to_q.weight",
851
+ "transformer_blocks.24.audio_attn1.to_k.weight",
852
+ "transformer_blocks.24.audio_attn1.to_out.0.weight",
853
+ "transformer_blocks.24.audio_attn1.to_q.weight",
854
+ "transformer_blocks.24.audio_attn1.to_v.weight",
855
+ "transformer_blocks.24.audio_attn2.to_k.weight",
856
+ "transformer_blocks.24.audio_attn2.to_out.0.weight",
857
+ "transformer_blocks.24.audio_attn2.to_q.weight",
858
+ "transformer_blocks.24.audio_attn2.to_v.weight",
859
+ "transformer_blocks.24.audio_to_video_attn.to_k.weight",
860
+ "transformer_blocks.24.audio_to_video_attn.to_out.0.weight",
861
+ "transformer_blocks.24.audio_to_video_attn.to_q.weight",
862
+ "transformer_blocks.24.audio_to_video_attn.to_v.weight",
863
+ "transformer_blocks.24.ff.net.0.proj.weight",
864
+ "transformer_blocks.24.ff.net.2.weight",
865
+ "transformer_blocks.24.video_to_audio_attn.to_k.weight",
866
+ "transformer_blocks.24.video_to_audio_attn.to_out.0.weight",
867
+ "transformer_blocks.24.video_to_audio_attn.to_q.weight",
868
+ "transformer_blocks.24.video_to_audio_attn.to_v.weight",
869
+ "transformer_blocks.24.audio_ff.net.0.proj.weight",
870
+ "transformer_blocks.24.audio_ff.net.2.weight",
871
+ "transformer_blocks.25.attn1.to_k.weight",
872
+ "transformer_blocks.25.attn1.to_out.0.weight",
873
+ "transformer_blocks.25.attn1.to_q.weight",
874
+ "transformer_blocks.25.attn1.to_v.weight",
875
+ "transformer_blocks.25.attn2.to_out.0.weight",
876
+ "transformer_blocks.25.attn2.to_q.weight",
877
+ "transformer_blocks.25.audio_attn1.to_k.weight",
878
+ "transformer_blocks.25.audio_attn1.to_out.0.weight",
879
+ "transformer_blocks.25.audio_attn1.to_q.weight",
880
+ "transformer_blocks.25.audio_attn2.to_k.weight",
881
+ "transformer_blocks.25.audio_attn2.to_out.0.weight",
882
+ "transformer_blocks.25.audio_attn2.to_q.weight",
883
+ "transformer_blocks.25.audio_attn2.to_v.weight",
884
+ "transformer_blocks.25.audio_ff.net.0.proj.weight",
885
+ "transformer_blocks.25.audio_ff.net.2.weight",
886
+ "transformer_blocks.25.audio_to_video_attn.to_k.weight",
887
+ "transformer_blocks.25.audio_to_video_attn.to_out.0.weight",
888
+ "transformer_blocks.25.audio_to_video_attn.to_q.weight",
889
+ "transformer_blocks.25.ff.net.0.proj.weight",
890
+ "transformer_blocks.25.ff.net.2.weight",
891
+ "transformer_blocks.25.video_to_audio_attn.to_k.weight",
892
+ "transformer_blocks.25.video_to_audio_attn.to_out.0.weight",
893
+ "transformer_blocks.25.video_to_audio_attn.to_q.weight",
894
+ "transformer_blocks.25.video_to_audio_attn.to_v.weight",
895
+ "transformer_blocks.26.attn1.to_k.weight",
896
+ "transformer_blocks.26.attn1.to_out.0.weight",
897
+ "transformer_blocks.26.attn1.to_q.weight",
898
+ "transformer_blocks.26.attn1.to_v.weight",
899
+ "transformer_blocks.26.attn2.to_out.0.weight",
900
+ "transformer_blocks.26.attn2.to_q.weight",
901
+ "transformer_blocks.26.audio_attn1.to_k.weight",
902
+ "transformer_blocks.26.audio_attn1.to_out.0.weight",
903
+ "transformer_blocks.26.audio_attn1.to_q.weight",
904
+ "transformer_blocks.26.audio_attn1.to_v.weight",
905
+ "transformer_blocks.26.audio_attn2.to_k.weight",
906
+ "transformer_blocks.26.audio_attn2.to_out.0.weight",
907
+ "transformer_blocks.26.audio_attn2.to_q.weight",
908
+ "transformer_blocks.26.audio_attn2.to_v.weight",
909
+ "transformer_blocks.26.audio_ff.net.0.proj.weight",
910
+ "transformer_blocks.26.audio_ff.net.2.weight",
911
+ "transformer_blocks.26.audio_to_video_attn.to_k.weight",
912
+ "transformer_blocks.26.audio_to_video_attn.to_out.0.weight",
913
+ "transformer_blocks.26.audio_to_video_attn.to_q.weight",
914
+ "transformer_blocks.26.audio_to_video_attn.to_v.weight",
915
+ "transformer_blocks.26.ff.net.0.proj.weight",
916
+ "transformer_blocks.26.ff.net.2.weight",
917
+ "transformer_blocks.26.video_to_audio_attn.to_k.weight",
918
+ "transformer_blocks.26.video_to_audio_attn.to_out.0.weight",
919
+ "transformer_blocks.26.video_to_audio_attn.to_q.weight",
920
+ "transformer_blocks.26.video_to_audio_attn.to_v.weight",
921
+ "transformer_blocks.27.attn1.to_k.weight",
922
+ "transformer_blocks.27.attn1.to_out.0.weight",
923
+ "transformer_blocks.27.attn1.to_q.weight",
924
+ "transformer_blocks.27.attn1.to_v.weight",
925
+ "transformer_blocks.27.attn2.to_out.0.weight",
926
+ "transformer_blocks.27.attn2.to_q.weight",
927
+ "transformer_blocks.27.audio_attn1.to_k.weight",
928
+ "transformer_blocks.27.audio_attn1.to_out.0.weight",
929
+ "transformer_blocks.27.audio_attn1.to_q.weight",
930
+ "transformer_blocks.27.audio_attn2.to_k.weight",
931
+ "transformer_blocks.27.audio_attn2.to_out.0.weight",
932
+ "transformer_blocks.27.audio_attn2.to_q.weight",
933
+ "transformer_blocks.27.audio_attn2.to_v.weight",
934
+ "transformer_blocks.27.audio_ff.net.0.proj.weight",
935
+ "transformer_blocks.27.audio_ff.net.2.weight",
936
+ "transformer_blocks.27.audio_to_video_attn.to_k.weight",
937
+ "transformer_blocks.27.audio_to_video_attn.to_out.0.weight",
938
+ "transformer_blocks.27.audio_to_video_attn.to_q.weight",
939
+ "transformer_blocks.27.ff.net.0.proj.weight",
940
+ "transformer_blocks.27.ff.net.2.weight",
941
+ "transformer_blocks.27.video_to_audio_attn.to_k.weight",
942
+ "transformer_blocks.27.video_to_audio_attn.to_out.0.weight",
943
+ "transformer_blocks.27.video_to_audio_attn.to_q.weight",
944
+ "transformer_blocks.27.video_to_audio_attn.to_v.weight",
945
+ "transformer_blocks.28.attn1.to_k.weight",
946
+ "transformer_blocks.28.attn1.to_out.0.weight",
947
+ "transformer_blocks.28.attn1.to_q.weight",
948
+ "transformer_blocks.28.attn1.to_v.weight",
949
+ "transformer_blocks.28.attn2.to_out.0.weight",
950
+ "transformer_blocks.28.attn2.to_q.weight",
951
+ "transformer_blocks.28.attn2.to_v.weight",
952
+ "transformer_blocks.28.audio_attn1.to_k.weight",
953
+ "transformer_blocks.28.audio_attn1.to_out.0.weight",
954
+ "transformer_blocks.28.audio_attn1.to_q.weight",
955
+ "transformer_blocks.28.audio_attn2.to_k.weight",
956
+ "transformer_blocks.28.audio_attn2.to_out.0.weight",
957
+ "transformer_blocks.28.audio_attn2.to_q.weight",
958
+ "transformer_blocks.28.audio_attn2.to_v.weight",
959
+ "transformer_blocks.28.audio_ff.net.0.proj.weight",
960
+ "transformer_blocks.28.audio_ff.net.2.weight",
961
+ "transformer_blocks.28.audio_to_video_attn.to_k.weight",
962
+ "transformer_blocks.28.audio_to_video_attn.to_out.0.weight",
963
+ "transformer_blocks.28.audio_to_video_attn.to_q.weight",
964
+ "transformer_blocks.28.audio_to_video_attn.to_v.weight",
965
+ "transformer_blocks.28.ff.net.0.proj.weight",
966
+ "transformer_blocks.28.ff.net.2.weight",
967
+ "transformer_blocks.28.video_to_audio_attn.to_k.weight",
968
+ "transformer_blocks.28.video_to_audio_attn.to_out.0.weight",
969
+ "transformer_blocks.28.video_to_audio_attn.to_q.weight",
970
+ "transformer_blocks.28.video_to_audio_attn.to_v.weight",
971
+ "transformer_blocks.29.attn1.to_out.0.weight",
972
+ "transformer_blocks.29.attn1.to_q.weight",
973
+ "transformer_blocks.29.attn1.to_v.weight",
974
+ "transformer_blocks.29.attn2.to_out.0.weight",
975
+ "transformer_blocks.29.audio_attn1.to_k.weight",
976
+ "transformer_blocks.29.audio_attn1.to_out.0.weight",
977
+ "transformer_blocks.29.audio_attn1.to_q.weight",
978
+ "transformer_blocks.29.audio_attn1.to_v.weight",
979
+ "transformer_blocks.29.audio_attn2.to_k.weight",
980
+ "transformer_blocks.29.audio_attn2.to_out.0.weight",
981
+ "transformer_blocks.29.audio_attn2.to_q.weight",
982
+ "transformer_blocks.29.audio_attn2.to_v.weight",
983
+ "transformer_blocks.29.audio_ff.net.0.proj.weight",
984
+ "transformer_blocks.29.audio_ff.net.2.weight",
985
+ "transformer_blocks.29.audio_to_video_attn.to_k.weight",
986
+ "transformer_blocks.29.audio_to_video_attn.to_out.0.weight",
987
+ "transformer_blocks.29.audio_to_video_attn.to_q.weight",
988
+ "transformer_blocks.29.audio_to_video_attn.to_v.weight",
989
+ "transformer_blocks.29.ff.net.0.proj.weight",
990
+ "transformer_blocks.29.ff.net.2.weight",
991
+ "transformer_blocks.29.video_to_audio_attn.to_k.weight",
992
+ "transformer_blocks.29.video_to_audio_attn.to_out.0.weight",
993
+ "transformer_blocks.29.video_to_audio_attn.to_q.weight",
994
+ "transformer_blocks.30.attn1.to_out.0.weight",
995
+ "transformer_blocks.30.attn1.to_q.weight",
996
+ "transformer_blocks.30.attn1.to_v.weight",
997
+ "transformer_blocks.30.attn2.to_out.0.weight",
998
+ "transformer_blocks.30.attn2.to_q.weight",
999
+ "transformer_blocks.30.audio_attn1.to_k.weight",
1000
+ "transformer_blocks.30.audio_attn1.to_out.0.weight",
1001
+ "transformer_blocks.30.audio_attn1.to_q.weight",
1002
+ "transformer_blocks.30.audio_attn1.to_v.weight",
1003
+ "transformer_blocks.30.audio_attn2.to_k.weight",
1004
+ "transformer_blocks.30.audio_attn2.to_out.0.weight",
1005
+ "transformer_blocks.30.audio_attn2.to_q.weight",
1006
+ "transformer_blocks.30.audio_attn2.to_v.weight",
1007
+ "transformer_blocks.30.audio_ff.net.0.proj.weight",
1008
+ "transformer_blocks.30.audio_ff.net.2.weight",
1009
+ "transformer_blocks.30.audio_to_video_attn.to_k.weight",
1010
+ "transformer_blocks.30.audio_to_video_attn.to_out.0.weight",
1011
+ "transformer_blocks.30.audio_to_video_attn.to_q.weight",
1012
+ "transformer_blocks.30.audio_to_video_attn.to_v.weight",
1013
+ "transformer_blocks.30.ff.net.0.proj.weight",
1014
+ "transformer_blocks.30.ff.net.2.weight",
1015
+ "transformer_blocks.30.video_to_audio_attn.to_k.weight",
1016
+ "transformer_blocks.30.video_to_audio_attn.to_out.0.weight",
1017
+ "transformer_blocks.30.video_to_audio_attn.to_q.weight",
1018
+ "transformer_blocks.30.video_to_audio_attn.to_v.weight",
1019
+ "transformer_blocks.31.attn1.to_out.0.weight",
1020
+ "transformer_blocks.31.attn1.to_q.weight",
1021
+ "transformer_blocks.31.attn1.to_v.weight",
1022
+ "transformer_blocks.31.attn2.to_q.weight",
1023
+ "transformer_blocks.31.audio_attn1.to_k.weight",
1024
+ "transformer_blocks.31.audio_attn1.to_out.0.weight",
1025
+ "transformer_blocks.31.audio_attn1.to_q.weight",
1026
+ "transformer_blocks.31.audio_attn1.to_v.weight",
1027
+ "transformer_blocks.31.attn2.to_out.0.weight",
1028
+ "transformer_blocks.31.audio_attn2.to_k.weight",
1029
+ "transformer_blocks.31.audio_attn2.to_out.0.weight",
1030
+ "transformer_blocks.31.audio_attn2.to_q.weight",
1031
+ "transformer_blocks.31.audio_attn2.to_v.weight",
1032
+ "transformer_blocks.31.audio_ff.net.0.proj.weight",
1033
+ "transformer_blocks.31.audio_ff.net.2.weight",
1034
+ "transformer_blocks.31.audio_to_video_attn.to_k.weight",
1035
+ "transformer_blocks.31.audio_to_video_attn.to_out.0.weight",
1036
+ "transformer_blocks.31.audio_to_video_attn.to_q.weight",
1037
+ "transformer_blocks.31.audio_to_video_attn.to_v.weight",
1038
+ "transformer_blocks.31.ff.net.0.proj.weight",
1039
+ "transformer_blocks.31.ff.net.2.weight",
1040
+ "transformer_blocks.31.video_to_audio_attn.to_k.weight",
1041
+ "transformer_blocks.31.video_to_audio_attn.to_out.0.weight",
1042
+ "transformer_blocks.31.video_to_audio_attn.to_q.weight",
1043
+ "transformer_blocks.31.video_to_audio_attn.to_v.weight",
1044
+ "transformer_blocks.32.attn1.to_k.weight",
1045
+ "transformer_blocks.32.attn1.to_out.0.weight",
1046
+ "transformer_blocks.32.attn1.to_q.weight",
1047
+ "transformer_blocks.32.attn1.to_v.weight",
1048
+ "transformer_blocks.32.attn2.to_out.0.weight",
1049
+ "transformer_blocks.32.attn2.to_q.weight",
1050
+ "transformer_blocks.32.audio_attn1.to_k.weight",
1051
+ "transformer_blocks.32.audio_attn1.to_out.0.weight",
1052
+ "transformer_blocks.32.audio_attn1.to_q.weight",
1053
+ "transformer_blocks.32.audio_attn1.to_v.weight",
1054
+ "transformer_blocks.32.audio_attn2.to_k.weight",
1055
+ "transformer_blocks.32.audio_attn2.to_out.0.weight",
1056
+ "transformer_blocks.32.audio_attn2.to_q.weight",
1057
+ "transformer_blocks.32.audio_attn2.to_v.weight",
1058
+ "transformer_blocks.32.audio_ff.net.0.proj.weight",
1059
+ "transformer_blocks.32.audio_ff.net.2.weight",
1060
+ "transformer_blocks.32.audio_to_video_attn.to_k.weight",
1061
+ "transformer_blocks.32.audio_to_video_attn.to_out.0.weight",
1062
+ "transformer_blocks.32.audio_to_video_attn.to_q.weight",
1063
+ "transformer_blocks.32.audio_to_video_attn.to_v.weight",
1064
+ "transformer_blocks.32.ff.net.0.proj.weight",
1065
+ "transformer_blocks.32.ff.net.2.weight",
1066
+ "transformer_blocks.32.video_to_audio_attn.to_k.weight",
1067
+ "transformer_blocks.32.video_to_audio_attn.to_out.0.weight",
1068
+ "transformer_blocks.32.video_to_audio_attn.to_q.weight",
1069
+ "transformer_blocks.33.attn1.to_out.0.weight",
1070
+ "transformer_blocks.33.attn1.to_q.weight",
1071
+ "transformer_blocks.33.attn1.to_v.weight",
1072
+ "transformer_blocks.33.attn2.to_out.0.weight",
1073
+ "transformer_blocks.33.attn2.to_q.weight",
1074
+ "transformer_blocks.33.audio_attn1.to_k.weight",
1075
+ "transformer_blocks.33.audio_attn1.to_out.0.weight",
1076
+ "transformer_blocks.33.audio_attn1.to_q.weight",
1077
+ "transformer_blocks.33.audio_attn1.to_v.weight",
1078
+ "transformer_blocks.33.audio_attn2.to_k.weight",
1079
+ "transformer_blocks.33.audio_attn2.to_out.0.weight",
1080
+ "transformer_blocks.33.audio_attn2.to_q.weight",
1081
+ "transformer_blocks.33.audio_attn2.to_v.weight",
1082
+ "transformer_blocks.33.audio_ff.net.0.proj.weight",
1083
+ "transformer_blocks.33.audio_ff.net.2.weight",
1084
+ "transformer_blocks.33.audio_to_video_attn.to_k.weight",
1085
+ "transformer_blocks.33.audio_to_video_attn.to_out.0.weight",
1086
+ "transformer_blocks.33.audio_to_video_attn.to_q.weight",
1087
+ "transformer_blocks.33.audio_to_video_attn.to_v.weight",
1088
+ "transformer_blocks.33.ff.net.0.proj.weight",
1089
+ "transformer_blocks.33.ff.net.2.weight",
1090
+ "transformer_blocks.33.video_to_audio_attn.to_k.weight",
1091
+ "transformer_blocks.33.video_to_audio_attn.to_out.0.weight",
1092
+ "transformer_blocks.33.video_to_audio_attn.to_q.weight",
1093
+ "transformer_blocks.33.video_to_audio_attn.to_v.weight",
1094
+ "transformer_blocks.34.attn1.to_k.weight",
1095
+ "transformer_blocks.34.attn1.to_out.0.weight",
1096
+ "transformer_blocks.34.attn1.to_q.weight",
1097
+ "transformer_blocks.34.attn1.to_v.weight",
1098
+ "transformer_blocks.34.attn2.to_q.weight",
1099
+ "transformer_blocks.34.audio_attn1.to_k.weight",
1100
+ "transformer_blocks.34.audio_attn1.to_out.0.weight",
1101
+ "transformer_blocks.34.audio_attn1.to_q.weight",
1102
+ "transformer_blocks.34.audio_attn1.to_v.weight",
1103
+ "transformer_blocks.34.audio_attn2.to_k.weight",
1104
+ "transformer_blocks.34.audio_attn2.to_out.0.weight",
1105
+ "transformer_blocks.34.audio_attn2.to_q.weight",
1106
+ "transformer_blocks.34.audio_attn2.to_v.weight",
1107
+ "transformer_blocks.34.audio_ff.net.0.proj.weight",
1108
+ "transformer_blocks.34.audio_ff.net.2.weight",
1109
+ "transformer_blocks.34.audio_to_video_attn.to_k.weight",
1110
+ "transformer_blocks.34.audio_to_video_attn.to_out.0.weight",
1111
+ "transformer_blocks.34.audio_to_video_attn.to_q.weight",
1112
+ "transformer_blocks.34.audio_to_video_attn.to_v.weight",
1113
+ "transformer_blocks.34.ff.net.0.proj.weight",
1114
+ "transformer_blocks.34.ff.net.2.weight",
1115
+ "transformer_blocks.34.video_to_audio_attn.to_k.weight",
1116
+ "transformer_blocks.34.video_to_audio_attn.to_out.0.weight",
1117
+ "transformer_blocks.34.video_to_audio_attn.to_q.weight",
1118
+ "transformer_blocks.34.video_to_audio_attn.to_v.weight",
1119
+ "transformer_blocks.35.attn1.to_k.weight",
1120
+ "transformer_blocks.35.attn1.to_out.0.weight",
1121
+ "transformer_blocks.35.attn1.to_q.weight",
1122
+ "transformer_blocks.35.attn1.to_v.weight",
1123
+ "transformer_blocks.35.attn2.to_q.weight",
1124
+ "transformer_blocks.35.audio_attn1.to_k.weight",
1125
+ "transformer_blocks.35.audio_attn1.to_out.0.weight",
1126
+ "transformer_blocks.35.audio_attn1.to_q.weight",
1127
+ "transformer_blocks.35.audio_attn1.to_v.weight",
1128
+ "transformer_blocks.35.audio_attn2.to_k.weight",
1129
+ "transformer_blocks.35.audio_attn2.to_out.0.weight",
1130
+ "transformer_blocks.35.audio_attn2.to_q.weight",
1131
+ "transformer_blocks.35.audio_attn2.to_v.weight",
1132
+ "transformer_blocks.35.audio_ff.net.0.proj.weight",
1133
+ "transformer_blocks.35.audio_ff.net.2.weight",
1134
+ "transformer_blocks.35.audio_to_video_attn.to_k.weight",
1135
+ "transformer_blocks.35.audio_to_video_attn.to_out.0.weight",
1136
+ "transformer_blocks.35.audio_to_video_attn.to_q.weight",
1137
+ "transformer_blocks.35.audio_to_video_attn.to_v.weight",
1138
+ "transformer_blocks.35.ff.net.0.proj.weight",
1139
+ "transformer_blocks.35.ff.net.2.weight",
1140
+ "transformer_blocks.35.video_to_audio_attn.to_k.weight",
1141
+ "transformer_blocks.35.video_to_audio_attn.to_out.0.weight",
1142
+ "transformer_blocks.35.video_to_audio_attn.to_q.weight",
1143
+ "transformer_blocks.35.video_to_audio_attn.to_v.weight",
1144
+ "transformer_blocks.36.attn1.to_k.weight",
1145
+ "transformer_blocks.36.attn1.to_out.0.weight",
1146
+ "transformer_blocks.36.attn1.to_q.weight",
1147
+ "transformer_blocks.36.attn1.to_v.weight",
1148
+ "transformer_blocks.36.attn2.to_q.weight",
1149
+ "transformer_blocks.36.audio_attn1.to_k.weight",
1150
+ "transformer_blocks.36.audio_attn1.to_out.0.weight",
1151
+ "transformer_blocks.36.audio_attn1.to_q.weight",
1152
+ "transformer_blocks.36.audio_attn2.to_k.weight",
1153
+ "transformer_blocks.36.audio_attn2.to_out.0.weight",
1154
+ "transformer_blocks.36.audio_attn2.to_q.weight",
1155
+ "transformer_blocks.36.audio_attn2.to_v.weight",
1156
+ "transformer_blocks.36.audio_ff.net.0.proj.weight",
1157
+ "transformer_blocks.36.audio_ff.net.2.weight",
1158
+ "transformer_blocks.36.audio_to_video_attn.to_k.weight",
1159
+ "transformer_blocks.36.audio_to_video_attn.to_out.0.weight",
1160
+ "transformer_blocks.36.audio_to_video_attn.to_q.weight",
1161
+ "transformer_blocks.36.ff.net.0.proj.weight",
1162
+ "transformer_blocks.36.ff.net.2.weight",
1163
+ "transformer_blocks.36.video_to_audio_attn.to_k.weight",
1164
+ "transformer_blocks.36.video_to_audio_attn.to_out.0.weight",
1165
+ "transformer_blocks.36.video_to_audio_attn.to_q.weight",
1166
+ "transformer_blocks.36.video_to_audio_attn.to_v.weight",
1167
+ "transformer_blocks.37.attn1.to_out.0.weight",
1168
+ "transformer_blocks.37.attn1.to_q.weight",
1169
+ "transformer_blocks.37.attn2.to_q.weight",
1170
+ "transformer_blocks.37.audio_attn1.to_k.weight",
1171
+ "transformer_blocks.37.audio_attn1.to_out.0.weight",
1172
+ "transformer_blocks.37.audio_attn1.to_q.weight",
1173
+ "transformer_blocks.37.audio_attn1.to_v.weight",
1174
+ "transformer_blocks.37.audio_attn2.to_k.weight",
1175
+ "transformer_blocks.37.audio_attn2.to_out.0.weight",
1176
+ "transformer_blocks.37.audio_attn2.to_q.weight",
1177
+ "transformer_blocks.37.audio_attn2.to_v.weight",
1178
+ "transformer_blocks.37.audio_to_video_attn.to_k.weight",
1179
+ "transformer_blocks.37.audio_to_video_attn.to_out.0.weight",
1180
+ "transformer_blocks.37.audio_to_video_attn.to_q.weight",
1181
+ "transformer_blocks.37.audio_to_video_attn.to_v.weight",
1182
+ "transformer_blocks.37.ff.net.0.proj.weight",
1183
+ "transformer_blocks.37.video_to_audio_attn.to_k.weight",
1184
+ "transformer_blocks.37.video_to_audio_attn.to_out.0.weight",
1185
+ "transformer_blocks.37.video_to_audio_attn.to_q.weight",
1186
+ "transformer_blocks.37.video_to_audio_attn.to_v.weight",
1187
+ "transformer_blocks.37.audio_ff.net.0.proj.weight",
1188
+ "transformer_blocks.37.audio_ff.net.2.weight",
1189
+ "transformer_blocks.37.ff.net.2.weight",
1190
+ "transformer_blocks.38.attn1.to_out.0.weight",
1191
+ "transformer_blocks.38.attn1.to_q.weight",
1192
+ "transformer_blocks.38.attn2.to_q.weight",
1193
+ "transformer_blocks.38.audio_attn1.to_k.weight",
1194
+ "transformer_blocks.38.audio_attn1.to_out.0.weight",
1195
+ "transformer_blocks.38.audio_attn1.to_q.weight",
1196
+ "transformer_blocks.38.audio_attn2.to_k.weight",
1197
+ "transformer_blocks.38.audio_attn2.to_out.0.weight",
1198
+ "transformer_blocks.38.audio_attn2.to_q.weight",
1199
+ "transformer_blocks.38.audio_attn2.to_v.weight",
1200
+ "transformer_blocks.38.audio_ff.net.0.proj.weight",
1201
+ "transformer_blocks.38.audio_ff.net.2.weight",
1202
+ "transformer_blocks.38.audio_to_video_attn.to_k.weight",
1203
+ "transformer_blocks.38.audio_to_video_attn.to_out.0.weight",
1204
+ "transformer_blocks.38.audio_to_video_attn.to_q.weight",
1205
+ "transformer_blocks.38.ff.net.0.proj.weight",
1206
+ "transformer_blocks.38.ff.net.2.weight",
1207
+ "transformer_blocks.38.video_to_audio_attn.to_k.weight",
1208
+ "transformer_blocks.38.video_to_audio_attn.to_out.0.weight",
1209
+ "transformer_blocks.38.video_to_audio_attn.to_q.weight",
1210
+ "transformer_blocks.38.video_to_audio_attn.to_v.weight",
1211
+ "transformer_blocks.39.attn1.to_out.0.weight",
1212
+ "transformer_blocks.39.attn1.to_q.weight",
1213
+ "transformer_blocks.39.attn2.to_q.weight",
1214
+ "transformer_blocks.39.audio_attn1.to_k.weight",
1215
+ "transformer_blocks.39.audio_attn1.to_out.0.weight",
1216
+ "transformer_blocks.39.audio_attn1.to_q.weight",
1217
+ "transformer_blocks.39.audio_attn1.to_v.weight",
1218
+ "transformer_blocks.39.audio_attn2.to_k.weight",
1219
+ "transformer_blocks.39.audio_attn2.to_out.0.weight",
1220
+ "transformer_blocks.39.audio_attn2.to_q.weight",
1221
+ "transformer_blocks.39.audio_attn2.to_v.weight",
1222
+ "transformer_blocks.39.audio_ff.net.0.proj.weight",
1223
+ "transformer_blocks.39.audio_ff.net.2.weight",
1224
+ "transformer_blocks.39.audio_to_video_attn.to_k.weight",
1225
+ "transformer_blocks.39.audio_to_video_attn.to_out.0.weight",
1226
+ "transformer_blocks.39.audio_to_video_attn.to_q.weight",
1227
+ "transformer_blocks.39.audio_to_video_attn.to_v.weight",
1228
+ "transformer_blocks.39.ff.net.0.proj.weight",
1229
+ "transformer_blocks.39.ff.net.2.weight",
1230
+ "transformer_blocks.39.video_to_audio_attn.to_k.weight",
1231
+ "transformer_blocks.39.video_to_audio_attn.to_out.0.weight",
1232
+ "transformer_blocks.39.video_to_audio_attn.to_q.weight",
1233
+ "transformer_blocks.39.video_to_audio_attn.to_v.weight",
1234
+ "transformer_blocks.40.attn1.to_out.0.weight",
1235
+ "transformer_blocks.40.attn1.to_q.weight",
1236
+ "transformer_blocks.40.attn2.to_q.weight",
1237
+ "transformer_blocks.40.audio_attn1.to_k.weight",
1238
+ "transformer_blocks.40.audio_attn1.to_out.0.weight",
1239
+ "transformer_blocks.40.audio_attn1.to_q.weight",
1240
+ "transformer_blocks.40.audio_attn1.to_v.weight",
1241
+ "transformer_blocks.40.audio_attn2.to_k.weight",
1242
+ "transformer_blocks.40.audio_attn2.to_out.0.weight",
1243
+ "transformer_blocks.40.audio_attn2.to_q.weight",
1244
+ "transformer_blocks.40.audio_attn2.to_v.weight",
1245
+ "transformer_blocks.40.audio_ff.net.0.proj.weight",
1246
+ "transformer_blocks.40.audio_ff.net.2.weight",
1247
+ "transformer_blocks.40.audio_to_video_attn.to_k.weight",
1248
+ "transformer_blocks.40.audio_to_video_attn.to_out.0.weight",
1249
+ "transformer_blocks.40.audio_to_video_attn.to_q.weight",
1250
+ "transformer_blocks.40.audio_to_video_attn.to_v.weight",
1251
+ "transformer_blocks.40.ff.net.0.proj.weight",
1252
+ "transformer_blocks.40.ff.net.2.weight",
1253
+ "transformer_blocks.40.video_to_audio_attn.to_out.0.weight",
1254
+ "transformer_blocks.40.video_to_audio_attn.to_q.weight",
1255
+ "transformer_blocks.40.video_to_audio_attn.to_v.weight",
1256
+ "transformer_blocks.41.attn1.to_out.0.weight",
1257
+ "transformer_blocks.41.attn1.to_q.weight",
1258
+ "transformer_blocks.41.attn2.to_q.weight",
1259
+ "transformer_blocks.41.audio_attn1.to_k.weight",
1260
+ "transformer_blocks.41.audio_attn1.to_out.0.weight",
1261
+ "transformer_blocks.41.audio_attn1.to_q.weight",
1262
+ "transformer_blocks.41.audio_attn2.to_k.weight",
1263
+ "transformer_blocks.41.audio_attn2.to_out.0.weight",
1264
+ "transformer_blocks.41.audio_attn2.to_q.weight",
1265
+ "transformer_blocks.41.audio_attn2.to_v.weight",
1266
+ "transformer_blocks.41.audio_ff.net.0.proj.weight",
1267
+ "transformer_blocks.41.audio_ff.net.2.weight",
1268
+ "transformer_blocks.41.audio_to_video_attn.to_k.weight",
1269
+ "transformer_blocks.41.audio_to_video_attn.to_out.0.weight",
1270
+ "transformer_blocks.41.audio_to_video_attn.to_q.weight",
1271
+ "transformer_blocks.41.ff.net.0.proj.weight",
1272
+ "transformer_blocks.41.ff.net.2.weight",
1273
+ "transformer_blocks.41.video_to_audio_attn.to_out.0.weight",
1274
+ "transformer_blocks.41.video_to_audio_attn.to_q.weight",
1275
+ "transformer_blocks.41.video_to_audio_attn.to_v.weight",
1276
+ "transformer_blocks.42.attn1.to_out.0.weight",
1277
+ "transformer_blocks.42.attn1.to_q.weight",
1278
+ "transformer_blocks.42.attn2.to_q.weight",
1279
+ "transformer_blocks.42.audio_attn1.to_k.weight",
1280
+ "transformer_blocks.42.audio_attn1.to_out.0.weight",
1281
+ "transformer_blocks.42.audio_attn1.to_q.weight",
1282
+ "transformer_blocks.42.audio_attn1.to_v.weight",
1283
+ "transformer_blocks.42.audio_attn2.to_k.weight",
1284
+ "transformer_blocks.42.audio_attn2.to_out.0.weight",
1285
+ "transformer_blocks.42.audio_attn2.to_q.weight",
1286
+ "transformer_blocks.42.audio_attn2.to_v.weight",
1287
+ "transformer_blocks.42.audio_ff.net.0.proj.weight",
1288
+ "transformer_blocks.42.audio_to_video_attn.to_k.weight",
1289
+ "transformer_blocks.42.audio_to_video_attn.to_out.0.weight",
1290
+ "transformer_blocks.42.audio_to_video_attn.to_q.weight",
1291
+ "transformer_blocks.42.ff.net.0.proj.weight",
1292
+ "transformer_blocks.42.ff.net.2.weight",
1293
+ "transformer_blocks.42.video_to_audio_attn.to_out.0.weight",
1294
+ "transformer_blocks.42.video_to_audio_attn.to_q.weight",
1295
+ "transformer_blocks.42.video_to_audio_attn.to_v.weight",
1296
+ "transformer_blocks.43.attn1.to_out.0.weight",
1297
+ "transformer_blocks.43.attn1.to_q.weight",
1298
+ "transformer_blocks.43.attn1.to_v.weight",
1299
+ "transformer_blocks.43.attn2.to_q.weight",
1300
+ "transformer_blocks.43.audio_attn1.to_k.weight",
1301
+ "transformer_blocks.43.audio_attn1.to_out.0.weight",
1302
+ "transformer_blocks.43.audio_attn1.to_q.weight",
1303
+ "transformer_blocks.43.audio_attn1.to_v.weight",
1304
+ "transformer_blocks.43.audio_attn2.to_k.weight",
1305
+ "transformer_blocks.43.audio_attn2.to_out.0.weight",
1306
+ "transformer_blocks.43.audio_attn2.to_q.weight",
1307
+ "transformer_blocks.43.audio_attn2.to_v.weight",
1308
+ "transformer_blocks.43.audio_ff.net.0.proj.weight",
1309
+ "transformer_blocks.43.audio_to_video_attn.to_k.weight",
1310
+ "transformer_blocks.43.audio_to_video_attn.to_out.0.weight",
1311
+ "transformer_blocks.43.audio_to_video_attn.to_q.weight",
1312
+ "transformer_blocks.43.audio_to_video_attn.to_v.weight",
1313
+ "transformer_blocks.43.ff.net.0.proj.weight",
1314
+ "transformer_blocks.43.ff.net.2.weight",
1315
+ "transformer_blocks.43.video_to_audio_attn.to_out.0.weight",
1316
+ "transformer_blocks.43.video_to_audio_attn.to_q.weight",
1317
+ "transformer_blocks.43.video_to_audio_attn.to_v.weight",
1318
+ "transformer_blocks.44.attn1.to_out.0.weight",
1319
+ "transformer_blocks.44.attn1.to_q.weight",
1320
+ "transformer_blocks.44.attn1.to_v.weight",
1321
+ "transformer_blocks.44.audio_attn1.to_k.weight",
1322
+ "transformer_blocks.44.audio_attn1.to_q.weight",
1323
+ "transformer_blocks.44.audio_attn1.to_v.weight",
1324
+ "transformer_blocks.44.attn2.to_q.weight",
1325
+ "transformer_blocks.44.audio_attn1.to_out.0.weight",
1326
+ "transformer_blocks.44.audio_attn2.to_k.weight",
1327
+ "transformer_blocks.44.audio_attn2.to_out.0.weight",
1328
+ "transformer_blocks.44.audio_attn2.to_q.weight",
1329
+ "transformer_blocks.44.audio_attn2.to_v.weight",
1330
+ "transformer_blocks.44.audio_ff.net.0.proj.weight",
1331
+ "transformer_blocks.44.audio_to_video_attn.to_k.weight",
1332
+ "transformer_blocks.44.audio_to_video_attn.to_out.0.weight",
1333
+ "transformer_blocks.44.audio_to_video_attn.to_q.weight",
1334
+ "transformer_blocks.44.audio_to_video_attn.to_v.weight",
1335
+ "transformer_blocks.44.ff.net.0.proj.weight",
1336
+ "transformer_blocks.44.ff.net.2.weight",
1337
+ "transformer_blocks.44.video_to_audio_attn.to_out.0.weight",
1338
+ "transformer_blocks.44.video_to_audio_attn.to_q.weight",
1339
+ "transformer_blocks.44.video_to_audio_attn.to_v.weight",
1340
+ "transformer_blocks.45.attn1.to_out.0.weight",
1341
+ "transformer_blocks.45.attn1.to_q.weight",
1342
+ "transformer_blocks.45.attn2.to_q.weight",
1343
+ "transformer_blocks.45.audio_attn1.to_k.weight",
1344
+ "transformer_blocks.45.audio_attn1.to_out.0.weight",
1345
+ "transformer_blocks.45.audio_attn1.to_q.weight",
1346
+ "transformer_blocks.45.audio_attn1.to_v.weight",
1347
+ "transformer_blocks.45.audio_attn2.to_k.weight",
1348
+ "transformer_blocks.45.audio_attn2.to_out.0.weight",
1349
+ "transformer_blocks.45.audio_attn2.to_q.weight",
1350
+ "transformer_blocks.45.audio_attn2.to_v.weight",
1351
+ "transformer_blocks.45.audio_ff.net.0.proj.weight",
1352
+ "transformer_blocks.45.audio_to_video_attn.to_k.weight",
1353
+ "transformer_blocks.45.audio_to_video_attn.to_out.0.weight",
1354
+ "transformer_blocks.45.audio_to_video_attn.to_q.weight",
1355
+ "transformer_blocks.45.audio_to_video_attn.to_v.weight",
1356
+ "transformer_blocks.45.ff.net.0.proj.weight",
1357
+ "transformer_blocks.45.ff.net.2.weight",
1358
+ "transformer_blocks.45.video_to_audio_attn.to_out.0.weight",
1359
+ "transformer_blocks.45.video_to_audio_attn.to_q.weight",
1360
+ "transformer_blocks.45.video_to_audio_attn.to_v.weight",
1361
+ "transformer_blocks.46.attn1.to_out.0.weight",
1362
+ "transformer_blocks.46.attn1.to_q.weight",
1363
+ "transformer_blocks.46.attn2.to_q.weight",
1364
+ "transformer_blocks.46.audio_attn1.to_out.0.weight",
1365
+ "transformer_blocks.46.audio_attn1.to_q.weight",
1366
+ "transformer_blocks.46.audio_attn2.to_k.weight",
1367
+ "transformer_blocks.46.audio_attn2.to_out.0.weight",
1368
+ "transformer_blocks.46.audio_attn2.to_q.weight",
1369
+ "transformer_blocks.46.audio_attn2.to_v.weight",
1370
+ "transformer_blocks.46.audio_ff.net.0.proj.weight",
1371
+ "transformer_blocks.46.audio_to_video_attn.to_out.0.weight",
1372
+ "transformer_blocks.46.audio_to_video_attn.to_q.weight",
1373
+ "transformer_blocks.46.ff.net.0.proj.weight",
1374
+ "transformer_blocks.46.ff.net.2.weight",
1375
+ "transformer_blocks.46.video_to_audio_attn.to_out.0.weight",
1376
+ "transformer_blocks.46.video_to_audio_attn.to_q.weight",
1377
+ "transformer_blocks.46.video_to_audio_attn.to_v.weight",
1378
+ "transformer_blocks.47.attn1.to_out.0.weight",
1379
+ "transformer_blocks.47.attn2.to_k.weight",
1380
+ "transformer_blocks.47.attn2.to_q.weight",
1381
+ "transformer_blocks.47.audio_attn1.to_k.weight",
1382
+ "transformer_blocks.47.audio_attn1.to_out.0.weight",
1383
+ "transformer_blocks.47.audio_attn1.to_q.weight",
1384
+ "transformer_blocks.47.audio_attn2.to_k.weight",
1385
+ "transformer_blocks.47.audio_attn2.to_q.weight",
1386
+ "transformer_blocks.47.audio_attn2.to_v.weight",
1387
+ "transformer_blocks.47.audio_ff.net.0.proj.weight",
1388
+ "transformer_blocks.47.audio_to_video_attn.to_k.weight",
1389
+ "transformer_blocks.47.audio_to_video_attn.to_out.0.weight",
1390
+ "transformer_blocks.47.ff.net.0.proj.weight",
1391
+ "transformer_blocks.47.video_to_audio_attn.to_out.0.weight",
1392
+ "transformer_blocks.47.video_to_audio_attn.to_q.weight"
1393
+ ]
1394
+ },
1395
+ "modules_to_not_convert": [
1396
+ "proj_in",
1397
+ "av_cross_attn_video_scale_shift",
1398
+ "time_embed",
1399
+ "av_cross_attn_video_a2v_gate",
1400
+ "audio_time_embed",
1401
+ "audio_proj_out",
1402
+ "av_cross_attn_audio_v2a_gate",
1403
+ "proj_out",
1404
+ "av_cross_attn_audio_scale_shift",
1405
+ "audio_caption_projection",
1406
+ "audio_proj_in",
1407
+ "caption_projection"
1408
+ ],
1409
+ "non_blocking": false,
1410
+ "quant_conv": false,
1411
+ "quant_method": "sdnq",
1412
+ "quantization_device": "cuda",
1413
+ "quantized_matmul_dtype": "int8",
1414
+ "return_device": "cuda",
1415
+ "sdnq_version": "0.1.4",
1416
+ "svd_rank": 32,
1417
+ "svd_steps": 8,
1418
+ "use_dynamic_quantization": true,
1419
+ "use_grad_ckpt": true,
1420
+ "use_quantized_matmul": false,
1421
+ "use_quantized_matmul_conv": false,
1422
+ "use_static_quantization": true,
1423
+ "use_stochastic_rounding": false,
1424
+ "use_svd": false,
1425
+ "weights_dtype": "int4"
1426
+ },
1427
+ "rope_double_precision": true,
1428
+ "rope_theta": 10000.0,
1429
+ "rope_type": "split",
1430
+ "timestep_scale_multiplier": 1000,
1431
+ "vae_scale_factors": [
1432
+ 8,
1433
+ 32,
1434
+ 32
1435
+ ]
1436
+ }
transformer/diffusion_pytorch_model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d835f031d3a13be75256af375acf1b2634bfe9e1bf231f30b089a2bb2d6f5e49
3
+ size 4962727784
transformer/diffusion_pytorch_model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:704f3501e4299e066923474e4111178e4375f7ef8c95f65a3d56f3e19fb60e5a
3
+ size 4998728648
transformer/diffusion_pytorch_model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36dd43f65e405a10946919986e5cd180c6d873bb141533acd667998b081ebebb
3
+ size 1529567208
transformer/diffusion_pytorch_model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
transformer/quantization_config.json ADDED
@@ -0,0 +1,1391 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_skip_keys": false,
3
+ "dequantize_fp32": false,
4
+ "dynamic_loss_threshold": 0.01,
5
+ "group_size": 0,
6
+ "is_integer": true,
7
+ "is_training": false,
8
+ "modules_dtype_dict": {
9
+ "float5_e3m1fn": [
10
+ "transformer_blocks.0.attn1.to_k.weight",
11
+ "transformer_blocks.0.attn1.to_v.weight",
12
+ "transformer_blocks.0.ff.net.2.weight",
13
+ "transformer_blocks.0.video_to_audio_attn.to_k.weight",
14
+ "transformer_blocks.0.video_to_audio_attn.to_v.weight",
15
+ "transformer_blocks.1.ff.net.0.proj.weight",
16
+ "transformer_blocks.3.ff.net.2.weight"
17
+ ],
18
+ "int5": [
19
+ "transformer_blocks.0.attn1.to_q.weight",
20
+ "transformer_blocks.0.audio_attn1.to_k.weight",
21
+ "transformer_blocks.0.audio_attn1.to_out.0.weight",
22
+ "transformer_blocks.0.audio_attn1.to_q.weight",
23
+ "transformer_blocks.0.audio_attn1.to_v.weight",
24
+ "transformer_blocks.0.audio_attn2.to_out.0.weight",
25
+ "transformer_blocks.0.audio_attn2.to_q.weight",
26
+ "transformer_blocks.0.audio_ff.net.0.proj.weight",
27
+ "transformer_blocks.0.audio_to_video_attn.to_k.weight",
28
+ "transformer_blocks.0.audio_to_video_attn.to_q.weight",
29
+ "transformer_blocks.0.audio_to_video_attn.to_v.weight",
30
+ "transformer_blocks.0.ff.net.0.proj.weight",
31
+ "transformer_blocks.0.video_to_audio_attn.to_q.weight",
32
+ "transformer_blocks.1.attn1.to_k.weight",
33
+ "transformer_blocks.1.attn1.to_q.weight",
34
+ "transformer_blocks.1.attn1.to_v.weight",
35
+ "transformer_blocks.1.audio_attn1.to_k.weight",
36
+ "transformer_blocks.1.audio_attn1.to_q.weight",
37
+ "transformer_blocks.1.audio_attn1.to_v.weight",
38
+ "transformer_blocks.1.audio_attn2.to_q.weight",
39
+ "transformer_blocks.1.audio_ff.net.0.proj.weight",
40
+ "transformer_blocks.1.audio_ff.net.2.weight",
41
+ "transformer_blocks.1.audio_to_video_attn.to_k.weight",
42
+ "transformer_blocks.1.audio_to_video_attn.to_q.weight",
43
+ "transformer_blocks.1.audio_to_video_attn.to_v.weight",
44
+ "transformer_blocks.1.ff.net.2.weight",
45
+ "transformer_blocks.1.video_to_audio_attn.to_k.weight",
46
+ "transformer_blocks.1.video_to_audio_attn.to_q.weight",
47
+ "transformer_blocks.1.video_to_audio_attn.to_v.weight",
48
+ "transformer_blocks.2.attn1.to_k.weight",
49
+ "transformer_blocks.2.attn1.to_q.weight",
50
+ "transformer_blocks.2.attn1.to_v.weight",
51
+ "transformer_blocks.2.audio_attn1.to_k.weight",
52
+ "transformer_blocks.2.audio_attn1.to_q.weight",
53
+ "transformer_blocks.2.audio_attn1.to_v.weight",
54
+ "transformer_blocks.2.audio_attn2.to_q.weight",
55
+ "transformer_blocks.2.audio_ff.net.2.weight",
56
+ "transformer_blocks.2.audio_to_video_attn.to_k.weight",
57
+ "transformer_blocks.2.audio_to_video_attn.to_q.weight",
58
+ "transformer_blocks.2.audio_to_video_attn.to_v.weight",
59
+ "transformer_blocks.2.ff.net.0.proj.weight",
60
+ "transformer_blocks.2.ff.net.2.weight",
61
+ "transformer_blocks.2.video_to_audio_attn.to_k.weight",
62
+ "transformer_blocks.2.video_to_audio_attn.to_q.weight",
63
+ "transformer_blocks.2.video_to_audio_attn.to_v.weight",
64
+ "transformer_blocks.3.attn1.to_k.weight",
65
+ "transformer_blocks.3.attn1.to_v.weight",
66
+ "transformer_blocks.3.audio_attn1.to_k.weight",
67
+ "transformer_blocks.3.audio_attn1.to_v.weight",
68
+ "transformer_blocks.3.audio_attn2.to_q.weight",
69
+ "transformer_blocks.3.audio_ff.net.2.weight",
70
+ "transformer_blocks.3.audio_to_video_attn.to_k.weight",
71
+ "transformer_blocks.3.audio_to_video_attn.to_q.weight",
72
+ "transformer_blocks.3.audio_to_video_attn.to_v.weight",
73
+ "transformer_blocks.3.ff.net.0.proj.weight",
74
+ "transformer_blocks.3.video_to_audio_attn.to_k.weight",
75
+ "transformer_blocks.3.video_to_audio_attn.to_v.weight",
76
+ "transformer_blocks.4.attn1.to_v.weight",
77
+ "transformer_blocks.4.attn2.to_q.weight",
78
+ "transformer_blocks.4.audio_attn1.to_k.weight",
79
+ "transformer_blocks.4.audio_attn1.to_v.weight",
80
+ "transformer_blocks.4.audio_attn2.to_q.weight",
81
+ "transformer_blocks.4.audio_ff.net.2.weight",
82
+ "transformer_blocks.4.audio_to_video_attn.to_k.weight",
83
+ "transformer_blocks.4.audio_to_video_attn.to_v.weight",
84
+ "transformer_blocks.4.ff.net.0.proj.weight",
85
+ "transformer_blocks.4.ff.net.2.weight",
86
+ "transformer_blocks.4.video_to_audio_attn.to_k.weight",
87
+ "transformer_blocks.4.video_to_audio_attn.to_v.weight",
88
+ "transformer_blocks.5.attn1.to_k.weight",
89
+ "transformer_blocks.5.attn1.to_v.weight",
90
+ "transformer_blocks.5.audio_attn2.to_q.weight",
91
+ "transformer_blocks.5.audio_to_video_attn.to_q.weight",
92
+ "transformer_blocks.5.video_to_audio_attn.to_k.weight",
93
+ "transformer_blocks.5.video_to_audio_attn.to_v.weight",
94
+ "transformer_blocks.10.attn2.to_k.weight",
95
+ "transformer_blocks.10.attn2.to_q.weight",
96
+ "transformer_blocks.10.audio_ff.net.2.weight",
97
+ "transformer_blocks.10.ff.net.2.weight",
98
+ "transformer_blocks.11.attn2.to_k.weight",
99
+ "transformer_blocks.11.audio_ff.net.2.weight",
100
+ "transformer_blocks.11.ff.net.2.weight",
101
+ "transformer_blocks.5.audio_ff.net.2.weight",
102
+ "transformer_blocks.5.ff.net.2.weight",
103
+ "transformer_blocks.6.audio_attn2.to_q.weight",
104
+ "transformer_blocks.6.audio_ff.net.2.weight",
105
+ "transformer_blocks.6.ff.net.2.weight",
106
+ "transformer_blocks.7.attn2.to_k.weight",
107
+ "transformer_blocks.7.attn2.to_q.weight",
108
+ "transformer_blocks.7.audio_ff.net.2.weight",
109
+ "transformer_blocks.7.ff.net.2.weight",
110
+ "transformer_blocks.8.audio_attn1.to_v.weight",
111
+ "transformer_blocks.8.audio_ff.net.2.weight",
112
+ "transformer_blocks.8.ff.net.2.weight",
113
+ "transformer_blocks.9.attn1.to_v.weight",
114
+ "transformer_blocks.9.audio_ff.net.2.weight",
115
+ "transformer_blocks.9.ff.net.2.weight",
116
+ "transformer_blocks.12.attn1.to_v.weight",
117
+ "transformer_blocks.12.attn2.to_k.weight",
118
+ "transformer_blocks.12.audio_ff.net.2.weight",
119
+ "transformer_blocks.12.ff.net.2.weight",
120
+ "transformer_blocks.13.audio_ff.net.2.weight",
121
+ "transformer_blocks.13.ff.net.2.weight",
122
+ "transformer_blocks.14.audio_ff.net.2.weight",
123
+ "transformer_blocks.14.ff.net.2.weight",
124
+ "transformer_blocks.15.audio_ff.net.2.weight",
125
+ "transformer_blocks.15.ff.net.2.weight",
126
+ "transformer_blocks.16.audio_ff.net.2.weight",
127
+ "transformer_blocks.16.ff.net.2.weight",
128
+ "transformer_blocks.17.audio_ff.net.2.weight",
129
+ "transformer_blocks.17.ff.net.2.weight",
130
+ "transformer_blocks.18.attn2.to_k.weight",
131
+ "transformer_blocks.18.ff.net.2.weight",
132
+ "transformer_blocks.19.attn2.to_k.weight",
133
+ "transformer_blocks.19.ff.net.2.weight",
134
+ "transformer_blocks.20.attn2.to_k.weight",
135
+ "transformer_blocks.20.ff.net.2.weight",
136
+ "transformer_blocks.21.attn2.to_k.weight",
137
+ "transformer_blocks.22.attn2.to_k.weight",
138
+ "transformer_blocks.23.attn2.to_k.weight",
139
+ "transformer_blocks.23.attn2.to_v.weight",
140
+ "transformer_blocks.24.attn2.to_k.weight",
141
+ "transformer_blocks.24.attn2.to_v.weight",
142
+ "transformer_blocks.25.attn2.to_k.weight",
143
+ "transformer_blocks.25.attn2.to_v.weight",
144
+ "transformer_blocks.25.audio_attn1.to_v.weight",
145
+ "transformer_blocks.25.audio_to_video_attn.to_v.weight",
146
+ "transformer_blocks.26.attn2.to_k.weight",
147
+ "transformer_blocks.26.attn2.to_v.weight",
148
+ "transformer_blocks.27.attn2.to_k.weight",
149
+ "transformer_blocks.27.attn2.to_v.weight",
150
+ "transformer_blocks.27.audio_attn1.to_v.weight",
151
+ "transformer_blocks.27.audio_to_video_attn.to_v.weight",
152
+ "transformer_blocks.28.attn2.to_k.weight",
153
+ "transformer_blocks.28.audio_attn1.to_v.weight",
154
+ "transformer_blocks.29.attn1.to_k.weight",
155
+ "transformer_blocks.29.attn2.to_k.weight",
156
+ "transformer_blocks.29.attn2.to_q.weight",
157
+ "transformer_blocks.29.attn2.to_v.weight",
158
+ "transformer_blocks.29.video_to_audio_attn.to_v.weight",
159
+ "transformer_blocks.30.attn1.to_k.weight",
160
+ "transformer_blocks.30.attn2.to_k.weight",
161
+ "transformer_blocks.30.attn2.to_v.weight",
162
+ "transformer_blocks.31.attn1.to_k.weight",
163
+ "transformer_blocks.31.attn2.to_k.weight",
164
+ "transformer_blocks.31.attn2.to_v.weight",
165
+ "transformer_blocks.32.attn2.to_k.weight",
166
+ "transformer_blocks.32.attn2.to_v.weight",
167
+ "transformer_blocks.32.video_to_audio_attn.to_v.weight",
168
+ "transformer_blocks.33.attn1.to_k.weight",
169
+ "transformer_blocks.33.attn2.to_k.weight",
170
+ "transformer_blocks.33.attn2.to_v.weight",
171
+ "transformer_blocks.34.attn2.to_k.weight",
172
+ "transformer_blocks.34.attn2.to_out.0.weight",
173
+ "transformer_blocks.34.attn2.to_v.weight",
174
+ "transformer_blocks.35.attn2.to_k.weight",
175
+ "transformer_blocks.35.attn2.to_out.0.weight",
176
+ "transformer_blocks.35.attn2.to_v.weight",
177
+ "transformer_blocks.36.attn2.to_k.weight",
178
+ "transformer_blocks.36.attn2.to_out.0.weight",
179
+ "transformer_blocks.36.attn2.to_v.weight",
180
+ "transformer_blocks.36.audio_attn1.to_v.weight",
181
+ "transformer_blocks.36.audio_to_video_attn.to_v.weight",
182
+ "transformer_blocks.37.attn1.to_k.weight",
183
+ "transformer_blocks.37.attn1.to_v.weight",
184
+ "transformer_blocks.37.attn2.to_k.weight",
185
+ "transformer_blocks.37.attn2.to_out.0.weight",
186
+ "transformer_blocks.37.attn2.to_v.weight",
187
+ "transformer_blocks.38.attn1.to_k.weight",
188
+ "transformer_blocks.38.attn1.to_v.weight",
189
+ "transformer_blocks.38.attn2.to_k.weight",
190
+ "transformer_blocks.38.attn2.to_out.0.weight",
191
+ "transformer_blocks.38.attn2.to_v.weight",
192
+ "transformer_blocks.38.audio_attn1.to_v.weight",
193
+ "transformer_blocks.38.audio_to_video_attn.to_v.weight",
194
+ "transformer_blocks.39.attn1.to_k.weight",
195
+ "transformer_blocks.39.attn1.to_v.weight",
196
+ "transformer_blocks.39.attn2.to_k.weight",
197
+ "transformer_blocks.39.attn2.to_out.0.weight",
198
+ "transformer_blocks.39.attn2.to_v.weight",
199
+ "transformer_blocks.40.attn1.to_k.weight",
200
+ "transformer_blocks.40.attn1.to_v.weight",
201
+ "transformer_blocks.40.attn2.to_k.weight",
202
+ "transformer_blocks.40.attn2.to_out.0.weight",
203
+ "transformer_blocks.40.attn2.to_v.weight",
204
+ "transformer_blocks.40.video_to_audio_attn.to_k.weight",
205
+ "transformer_blocks.41.attn1.to_k.weight",
206
+ "transformer_blocks.41.attn1.to_v.weight",
207
+ "transformer_blocks.41.attn2.to_k.weight",
208
+ "transformer_blocks.41.attn2.to_out.0.weight",
209
+ "transformer_blocks.41.attn2.to_v.weight",
210
+ "transformer_blocks.41.audio_attn1.to_v.weight",
211
+ "transformer_blocks.41.audio_to_video_attn.to_v.weight",
212
+ "transformer_blocks.41.video_to_audio_attn.to_k.weight",
213
+ "transformer_blocks.42.attn1.to_k.weight",
214
+ "transformer_blocks.42.attn1.to_v.weight",
215
+ "transformer_blocks.42.attn2.to_k.weight",
216
+ "transformer_blocks.42.attn2.to_out.0.weight",
217
+ "transformer_blocks.42.attn2.to_v.weight",
218
+ "transformer_blocks.42.audio_ff.net.2.weight",
219
+ "transformer_blocks.42.audio_to_video_attn.to_v.weight",
220
+ "transformer_blocks.42.video_to_audio_attn.to_k.weight",
221
+ "transformer_blocks.43.attn1.to_k.weight",
222
+ "transformer_blocks.43.attn2.to_k.weight",
223
+ "transformer_blocks.43.attn2.to_out.0.weight",
224
+ "transformer_blocks.43.attn2.to_v.weight",
225
+ "transformer_blocks.43.audio_ff.net.2.weight",
226
+ "transformer_blocks.43.video_to_audio_attn.to_k.weight",
227
+ "transformer_blocks.44.attn1.to_k.weight",
228
+ "transformer_blocks.44.attn2.to_k.weight",
229
+ "transformer_blocks.44.attn2.to_out.0.weight",
230
+ "transformer_blocks.44.attn2.to_v.weight",
231
+ "transformer_blocks.44.audio_ff.net.2.weight",
232
+ "transformer_blocks.44.video_to_audio_attn.to_k.weight",
233
+ "transformer_blocks.45.attn1.to_k.weight",
234
+ "transformer_blocks.45.attn1.to_v.weight",
235
+ "transformer_blocks.45.attn2.to_k.weight",
236
+ "transformer_blocks.45.attn2.to_out.0.weight",
237
+ "transformer_blocks.45.attn2.to_v.weight",
238
+ "transformer_blocks.45.audio_ff.net.2.weight",
239
+ "transformer_blocks.45.video_to_audio_attn.to_k.weight",
240
+ "transformer_blocks.46.attn1.to_k.weight",
241
+ "transformer_blocks.46.attn1.to_v.weight",
242
+ "transformer_blocks.46.attn2.to_k.weight",
243
+ "transformer_blocks.46.attn2.to_out.0.weight",
244
+ "transformer_blocks.46.attn2.to_v.weight",
245
+ "transformer_blocks.46.audio_attn1.to_k.weight",
246
+ "transformer_blocks.46.audio_attn1.to_v.weight",
247
+ "transformer_blocks.46.audio_ff.net.2.weight",
248
+ "transformer_blocks.46.audio_to_video_attn.to_k.weight",
249
+ "transformer_blocks.46.audio_to_video_attn.to_v.weight",
250
+ "transformer_blocks.46.video_to_audio_attn.to_k.weight",
251
+ "transformer_blocks.47.attn1.to_k.weight",
252
+ "transformer_blocks.47.attn1.to_q.weight",
253
+ "transformer_blocks.47.attn1.to_v.weight",
254
+ "transformer_blocks.47.attn2.to_out.0.weight",
255
+ "transformer_blocks.47.attn2.to_v.weight",
256
+ "transformer_blocks.47.audio_attn1.to_v.weight",
257
+ "transformer_blocks.47.audio_attn2.to_out.0.weight",
258
+ "transformer_blocks.47.audio_ff.net.2.weight",
259
+ "transformer_blocks.47.audio_to_video_attn.to_q.weight",
260
+ "transformer_blocks.47.audio_to_video_attn.to_v.weight",
261
+ "transformer_blocks.47.ff.net.2.weight",
262
+ "transformer_blocks.47.video_to_audio_attn.to_k.weight",
263
+ "transformer_blocks.47.video_to_audio_attn.to_v.weight"
264
+ ],
265
+ "uint4": [
266
+ "transformer_blocks.0.attn1.to_out.0.weight",
267
+ "transformer_blocks.0.attn2.to_k.weight",
268
+ "transformer_blocks.0.attn2.to_out.0.weight",
269
+ "transformer_blocks.0.attn2.to_q.weight",
270
+ "transformer_blocks.0.attn2.to_v.weight",
271
+ "transformer_blocks.0.audio_attn2.to_k.weight",
272
+ "transformer_blocks.0.audio_attn2.to_v.weight",
273
+ "transformer_blocks.0.audio_ff.net.2.weight",
274
+ "transformer_blocks.0.audio_to_video_attn.to_out.0.weight",
275
+ "transformer_blocks.0.video_to_audio_attn.to_out.0.weight",
276
+ "transformer_blocks.1.attn1.to_out.0.weight",
277
+ "transformer_blocks.1.attn2.to_k.weight",
278
+ "transformer_blocks.1.attn2.to_out.0.weight",
279
+ "transformer_blocks.1.attn2.to_q.weight",
280
+ "transformer_blocks.1.attn2.to_v.weight",
281
+ "transformer_blocks.1.audio_attn1.to_out.0.weight",
282
+ "transformer_blocks.1.audio_attn2.to_k.weight",
283
+ "transformer_blocks.1.audio_attn2.to_out.0.weight",
284
+ "transformer_blocks.1.audio_attn2.to_v.weight",
285
+ "transformer_blocks.1.audio_to_video_attn.to_out.0.weight",
286
+ "transformer_blocks.1.video_to_audio_attn.to_out.0.weight",
287
+ "transformer_blocks.2.attn1.to_out.0.weight",
288
+ "transformer_blocks.2.attn2.to_k.weight",
289
+ "transformer_blocks.2.attn2.to_out.0.weight",
290
+ "transformer_blocks.2.attn2.to_q.weight",
291
+ "transformer_blocks.2.attn2.to_v.weight",
292
+ "transformer_blocks.2.audio_attn1.to_out.0.weight",
293
+ "transformer_blocks.2.audio_attn2.to_k.weight",
294
+ "transformer_blocks.2.audio_attn2.to_out.0.weight",
295
+ "transformer_blocks.2.audio_attn2.to_v.weight",
296
+ "transformer_blocks.2.audio_ff.net.0.proj.weight",
297
+ "transformer_blocks.2.audio_to_video_attn.to_out.0.weight",
298
+ "transformer_blocks.2.video_to_audio_attn.to_out.0.weight",
299
+ "transformer_blocks.3.attn1.to_out.0.weight",
300
+ "transformer_blocks.3.attn1.to_q.weight",
301
+ "transformer_blocks.3.attn2.to_k.weight",
302
+ "transformer_blocks.3.attn2.to_out.0.weight",
303
+ "transformer_blocks.3.attn2.to_q.weight",
304
+ "transformer_blocks.3.attn2.to_v.weight",
305
+ "transformer_blocks.3.audio_attn1.to_out.0.weight",
306
+ "transformer_blocks.3.audio_attn1.to_q.weight",
307
+ "transformer_blocks.3.audio_attn2.to_k.weight",
308
+ "transformer_blocks.3.audio_attn2.to_out.0.weight",
309
+ "transformer_blocks.3.audio_attn2.to_v.weight",
310
+ "transformer_blocks.3.audio_ff.net.0.proj.weight",
311
+ "transformer_blocks.3.audio_to_video_attn.to_out.0.weight",
312
+ "transformer_blocks.3.video_to_audio_attn.to_out.0.weight",
313
+ "transformer_blocks.3.video_to_audio_attn.to_q.weight",
314
+ "transformer_blocks.4.attn1.to_k.weight",
315
+ "transformer_blocks.4.attn1.to_out.0.weight",
316
+ "transformer_blocks.4.attn1.to_q.weight",
317
+ "transformer_blocks.4.attn2.to_k.weight",
318
+ "transformer_blocks.4.attn2.to_out.0.weight",
319
+ "transformer_blocks.4.attn2.to_v.weight",
320
+ "transformer_blocks.4.audio_attn1.to_out.0.weight",
321
+ "transformer_blocks.4.audio_attn1.to_q.weight",
322
+ "transformer_blocks.4.audio_attn2.to_k.weight",
323
+ "transformer_blocks.4.audio_attn2.to_out.0.weight",
324
+ "transformer_blocks.4.audio_attn2.to_v.weight",
325
+ "transformer_blocks.4.audio_ff.net.0.proj.weight",
326
+ "transformer_blocks.4.audio_to_video_attn.to_out.0.weight",
327
+ "transformer_blocks.4.audio_to_video_attn.to_q.weight",
328
+ "transformer_blocks.4.video_to_audio_attn.to_out.0.weight",
329
+ "transformer_blocks.4.video_to_audio_attn.to_q.weight",
330
+ "transformer_blocks.5.attn1.to_out.0.weight",
331
+ "transformer_blocks.5.attn1.to_q.weight",
332
+ "transformer_blocks.5.attn2.to_k.weight",
333
+ "transformer_blocks.5.attn2.to_out.0.weight",
334
+ "transformer_blocks.5.attn2.to_q.weight",
335
+ "transformer_blocks.5.attn2.to_v.weight",
336
+ "transformer_blocks.5.audio_attn1.to_k.weight",
337
+ "transformer_blocks.5.audio_attn1.to_out.0.weight",
338
+ "transformer_blocks.5.audio_attn1.to_q.weight",
339
+ "transformer_blocks.5.audio_attn1.to_v.weight",
340
+ "transformer_blocks.5.audio_attn2.to_k.weight",
341
+ "transformer_blocks.5.audio_attn2.to_out.0.weight",
342
+ "transformer_blocks.5.audio_attn2.to_v.weight",
343
+ "transformer_blocks.5.audio_to_video_attn.to_k.weight",
344
+ "transformer_blocks.5.audio_to_video_attn.to_out.0.weight",
345
+ "transformer_blocks.5.audio_to_video_attn.to_v.weight",
346
+ "transformer_blocks.5.video_to_audio_attn.to_out.0.weight",
347
+ "transformer_blocks.5.video_to_audio_attn.to_q.weight",
348
+ "transformer_blocks.10.attn1.to_k.weight",
349
+ "transformer_blocks.10.attn1.to_out.0.weight",
350
+ "transformer_blocks.10.attn1.to_q.weight",
351
+ "transformer_blocks.10.attn1.to_v.weight",
352
+ "transformer_blocks.10.attn2.to_out.0.weight",
353
+ "transformer_blocks.10.attn2.to_v.weight",
354
+ "transformer_blocks.10.audio_attn1.to_k.weight",
355
+ "transformer_blocks.10.audio_attn1.to_out.0.weight",
356
+ "transformer_blocks.10.audio_attn1.to_q.weight",
357
+ "transformer_blocks.10.audio_attn1.to_v.weight",
358
+ "transformer_blocks.10.audio_attn2.to_k.weight",
359
+ "transformer_blocks.10.audio_attn2.to_out.0.weight",
360
+ "transformer_blocks.10.audio_attn2.to_q.weight",
361
+ "transformer_blocks.10.audio_attn2.to_v.weight",
362
+ "transformer_blocks.10.audio_ff.net.0.proj.weight",
363
+ "transformer_blocks.10.audio_to_video_attn.to_k.weight",
364
+ "transformer_blocks.10.audio_to_video_attn.to_out.0.weight",
365
+ "transformer_blocks.10.audio_to_video_attn.to_q.weight",
366
+ "transformer_blocks.10.audio_to_video_attn.to_v.weight",
367
+ "transformer_blocks.10.ff.net.0.proj.weight",
368
+ "transformer_blocks.10.video_to_audio_attn.to_k.weight",
369
+ "transformer_blocks.10.video_to_audio_attn.to_out.0.weight",
370
+ "transformer_blocks.10.video_to_audio_attn.to_q.weight",
371
+ "transformer_blocks.10.video_to_audio_attn.to_v.weight",
372
+ "transformer_blocks.11.attn1.to_k.weight",
373
+ "transformer_blocks.11.attn1.to_out.0.weight",
374
+ "transformer_blocks.11.attn1.to_q.weight",
375
+ "transformer_blocks.11.attn1.to_v.weight",
376
+ "transformer_blocks.11.attn2.to_out.0.weight",
377
+ "transformer_blocks.11.attn2.to_q.weight",
378
+ "transformer_blocks.11.attn2.to_v.weight",
379
+ "transformer_blocks.11.audio_attn1.to_k.weight",
380
+ "transformer_blocks.11.audio_attn1.to_out.0.weight",
381
+ "transformer_blocks.11.audio_attn1.to_q.weight",
382
+ "transformer_blocks.11.audio_attn1.to_v.weight",
383
+ "transformer_blocks.11.audio_attn2.to_k.weight",
384
+ "transformer_blocks.11.audio_attn2.to_out.0.weight",
385
+ "transformer_blocks.11.audio_attn2.to_q.weight",
386
+ "transformer_blocks.11.audio_attn2.to_v.weight",
387
+ "transformer_blocks.11.audio_ff.net.0.proj.weight",
388
+ "transformer_blocks.11.audio_to_video_attn.to_k.weight",
389
+ "transformer_blocks.11.audio_to_video_attn.to_out.0.weight",
390
+ "transformer_blocks.11.audio_to_video_attn.to_q.weight",
391
+ "transformer_blocks.11.audio_to_video_attn.to_v.weight",
392
+ "transformer_blocks.11.ff.net.0.proj.weight",
393
+ "transformer_blocks.11.video_to_audio_attn.to_k.weight",
394
+ "transformer_blocks.11.video_to_audio_attn.to_out.0.weight",
395
+ "transformer_blocks.11.video_to_audio_attn.to_q.weight",
396
+ "transformer_blocks.11.video_to_audio_attn.to_v.weight",
397
+ "transformer_blocks.5.audio_ff.net.0.proj.weight",
398
+ "transformer_blocks.5.ff.net.0.proj.weight",
399
+ "transformer_blocks.6.attn1.to_k.weight",
400
+ "transformer_blocks.6.attn1.to_out.0.weight",
401
+ "transformer_blocks.6.attn1.to_q.weight",
402
+ "transformer_blocks.6.attn1.to_v.weight",
403
+ "transformer_blocks.6.attn2.to_k.weight",
404
+ "transformer_blocks.6.attn2.to_out.0.weight",
405
+ "transformer_blocks.6.attn2.to_q.weight",
406
+ "transformer_blocks.6.attn2.to_v.weight",
407
+ "transformer_blocks.6.audio_attn1.to_k.weight",
408
+ "transformer_blocks.6.audio_attn1.to_out.0.weight",
409
+ "transformer_blocks.6.audio_attn1.to_q.weight",
410
+ "transformer_blocks.6.audio_attn1.to_v.weight",
411
+ "transformer_blocks.6.audio_attn2.to_k.weight",
412
+ "transformer_blocks.6.audio_attn2.to_out.0.weight",
413
+ "transformer_blocks.6.audio_attn2.to_v.weight",
414
+ "transformer_blocks.6.audio_ff.net.0.proj.weight",
415
+ "transformer_blocks.6.audio_to_video_attn.to_k.weight",
416
+ "transformer_blocks.6.audio_to_video_attn.to_out.0.weight",
417
+ "transformer_blocks.6.audio_to_video_attn.to_q.weight",
418
+ "transformer_blocks.6.audio_to_video_attn.to_v.weight",
419
+ "transformer_blocks.6.ff.net.0.proj.weight",
420
+ "transformer_blocks.6.video_to_audio_attn.to_k.weight",
421
+ "transformer_blocks.6.video_to_audio_attn.to_out.0.weight",
422
+ "transformer_blocks.6.video_to_audio_attn.to_q.weight",
423
+ "transformer_blocks.6.video_to_audio_attn.to_v.weight",
424
+ "transformer_blocks.7.attn1.to_k.weight",
425
+ "transformer_blocks.7.attn1.to_out.0.weight",
426
+ "transformer_blocks.7.attn1.to_q.weight",
427
+ "transformer_blocks.7.attn1.to_v.weight",
428
+ "transformer_blocks.7.attn2.to_out.0.weight",
429
+ "transformer_blocks.7.attn2.to_v.weight",
430
+ "transformer_blocks.7.audio_attn1.to_k.weight",
431
+ "transformer_blocks.7.audio_attn1.to_out.0.weight",
432
+ "transformer_blocks.7.audio_attn1.to_q.weight",
433
+ "transformer_blocks.7.audio_attn1.to_v.weight",
434
+ "transformer_blocks.7.audio_attn2.to_k.weight",
435
+ "transformer_blocks.7.audio_attn2.to_out.0.weight",
436
+ "transformer_blocks.7.audio_attn2.to_q.weight",
437
+ "transformer_blocks.7.audio_attn2.to_v.weight",
438
+ "transformer_blocks.7.audio_ff.net.0.proj.weight",
439
+ "transformer_blocks.7.audio_to_video_attn.to_k.weight",
440
+ "transformer_blocks.7.audio_to_video_attn.to_out.0.weight",
441
+ "transformer_blocks.7.audio_to_video_attn.to_q.weight",
442
+ "transformer_blocks.7.audio_to_video_attn.to_v.weight",
443
+ "transformer_blocks.7.ff.net.0.proj.weight",
444
+ "transformer_blocks.7.video_to_audio_attn.to_k.weight",
445
+ "transformer_blocks.7.video_to_audio_attn.to_out.0.weight",
446
+ "transformer_blocks.7.video_to_audio_attn.to_q.weight",
447
+ "transformer_blocks.7.video_to_audio_attn.to_v.weight",
448
+ "transformer_blocks.8.attn1.to_k.weight",
449
+ "transformer_blocks.8.attn1.to_out.0.weight",
450
+ "transformer_blocks.8.attn1.to_q.weight",
451
+ "transformer_blocks.8.attn1.to_v.weight",
452
+ "transformer_blocks.8.attn2.to_k.weight",
453
+ "transformer_blocks.8.attn2.to_out.0.weight",
454
+ "transformer_blocks.8.attn2.to_q.weight",
455
+ "transformer_blocks.8.attn2.to_v.weight",
456
+ "transformer_blocks.8.audio_attn1.to_k.weight",
457
+ "transformer_blocks.8.audio_attn1.to_out.0.weight",
458
+ "transformer_blocks.8.audio_attn1.to_q.weight",
459
+ "transformer_blocks.8.audio_attn2.to_k.weight",
460
+ "transformer_blocks.8.audio_attn2.to_out.0.weight",
461
+ "transformer_blocks.8.audio_attn2.to_q.weight",
462
+ "transformer_blocks.8.audio_attn2.to_v.weight",
463
+ "transformer_blocks.8.audio_ff.net.0.proj.weight",
464
+ "transformer_blocks.8.audio_to_video_attn.to_k.weight",
465
+ "transformer_blocks.8.audio_to_video_attn.to_out.0.weight",
466
+ "transformer_blocks.8.audio_to_video_attn.to_q.weight",
467
+ "transformer_blocks.8.audio_to_video_attn.to_v.weight",
468
+ "transformer_blocks.8.ff.net.0.proj.weight",
469
+ "transformer_blocks.8.video_to_audio_attn.to_k.weight",
470
+ "transformer_blocks.8.video_to_audio_attn.to_out.0.weight",
471
+ "transformer_blocks.8.video_to_audio_attn.to_q.weight",
472
+ "transformer_blocks.8.video_to_audio_attn.to_v.weight",
473
+ "transformer_blocks.9.attn1.to_k.weight",
474
+ "transformer_blocks.9.attn1.to_out.0.weight",
475
+ "transformer_blocks.9.attn1.to_q.weight",
476
+ "transformer_blocks.9.attn2.to_k.weight",
477
+ "transformer_blocks.9.attn2.to_out.0.weight",
478
+ "transformer_blocks.9.attn2.to_q.weight",
479
+ "transformer_blocks.9.attn2.to_v.weight",
480
+ "transformer_blocks.9.audio_attn1.to_k.weight",
481
+ "transformer_blocks.9.audio_attn1.to_out.0.weight",
482
+ "transformer_blocks.9.audio_attn1.to_q.weight",
483
+ "transformer_blocks.9.audio_attn1.to_v.weight",
484
+ "transformer_blocks.9.audio_attn2.to_k.weight",
485
+ "transformer_blocks.9.audio_attn2.to_out.0.weight",
486
+ "transformer_blocks.9.audio_attn2.to_q.weight",
487
+ "transformer_blocks.9.audio_attn2.to_v.weight",
488
+ "transformer_blocks.9.audio_ff.net.0.proj.weight",
489
+ "transformer_blocks.9.audio_to_video_attn.to_k.weight",
490
+ "transformer_blocks.9.audio_to_video_attn.to_out.0.weight",
491
+ "transformer_blocks.9.audio_to_video_attn.to_q.weight",
492
+ "transformer_blocks.9.audio_to_video_attn.to_v.weight",
493
+ "transformer_blocks.9.ff.net.0.proj.weight",
494
+ "transformer_blocks.9.video_to_audio_attn.to_k.weight",
495
+ "transformer_blocks.9.video_to_audio_attn.to_out.0.weight",
496
+ "transformer_blocks.9.video_to_audio_attn.to_q.weight",
497
+ "transformer_blocks.9.video_to_audio_attn.to_v.weight",
498
+ "transformer_blocks.12.attn1.to_k.weight",
499
+ "transformer_blocks.12.attn1.to_out.0.weight",
500
+ "transformer_blocks.12.attn1.to_q.weight",
501
+ "transformer_blocks.12.attn2.to_out.0.weight",
502
+ "transformer_blocks.12.attn2.to_q.weight",
503
+ "transformer_blocks.12.attn2.to_v.weight",
504
+ "transformer_blocks.12.audio_attn1.to_k.weight",
505
+ "transformer_blocks.12.audio_attn1.to_out.0.weight",
506
+ "transformer_blocks.12.audio_attn1.to_q.weight",
507
+ "transformer_blocks.12.audio_attn1.to_v.weight",
508
+ "transformer_blocks.12.audio_attn2.to_k.weight",
509
+ "transformer_blocks.12.audio_attn2.to_out.0.weight",
510
+ "transformer_blocks.12.audio_attn2.to_q.weight",
511
+ "transformer_blocks.12.audio_attn2.to_v.weight",
512
+ "transformer_blocks.12.audio_ff.net.0.proj.weight",
513
+ "transformer_blocks.12.audio_to_video_attn.to_k.weight",
514
+ "transformer_blocks.12.audio_to_video_attn.to_out.0.weight",
515
+ "transformer_blocks.12.audio_to_video_attn.to_q.weight",
516
+ "transformer_blocks.12.audio_to_video_attn.to_v.weight",
517
+ "transformer_blocks.12.ff.net.0.proj.weight",
518
+ "transformer_blocks.12.video_to_audio_attn.to_k.weight",
519
+ "transformer_blocks.12.video_to_audio_attn.to_out.0.weight",
520
+ "transformer_blocks.12.video_to_audio_attn.to_q.weight",
521
+ "transformer_blocks.12.video_to_audio_attn.to_v.weight",
522
+ "transformer_blocks.13.attn1.to_k.weight",
523
+ "transformer_blocks.13.attn1.to_out.0.weight",
524
+ "transformer_blocks.13.attn1.to_q.weight",
525
+ "transformer_blocks.13.attn1.to_v.weight",
526
+ "transformer_blocks.13.attn2.to_k.weight",
527
+ "transformer_blocks.13.attn2.to_out.0.weight",
528
+ "transformer_blocks.13.attn2.to_q.weight",
529
+ "transformer_blocks.13.attn2.to_v.weight",
530
+ "transformer_blocks.13.audio_attn1.to_k.weight",
531
+ "transformer_blocks.13.audio_attn1.to_out.0.weight",
532
+ "transformer_blocks.13.audio_attn1.to_q.weight",
533
+ "transformer_blocks.13.audio_attn1.to_v.weight",
534
+ "transformer_blocks.13.audio_attn2.to_k.weight",
535
+ "transformer_blocks.13.audio_attn2.to_out.0.weight",
536
+ "transformer_blocks.13.audio_attn2.to_q.weight",
537
+ "transformer_blocks.13.audio_attn2.to_v.weight",
538
+ "transformer_blocks.13.audio_ff.net.0.proj.weight",
539
+ "transformer_blocks.13.audio_to_video_attn.to_k.weight",
540
+ "transformer_blocks.13.audio_to_video_attn.to_out.0.weight",
541
+ "transformer_blocks.13.audio_to_video_attn.to_q.weight",
542
+ "transformer_blocks.13.audio_to_video_attn.to_v.weight",
543
+ "transformer_blocks.13.ff.net.0.proj.weight",
544
+ "transformer_blocks.13.video_to_audio_attn.to_k.weight",
545
+ "transformer_blocks.13.video_to_audio_attn.to_out.0.weight",
546
+ "transformer_blocks.13.video_to_audio_attn.to_q.weight",
547
+ "transformer_blocks.13.video_to_audio_attn.to_v.weight",
548
+ "transformer_blocks.14.attn1.to_k.weight",
549
+ "transformer_blocks.14.attn1.to_out.0.weight",
550
+ "transformer_blocks.14.attn1.to_q.weight",
551
+ "transformer_blocks.14.attn1.to_v.weight",
552
+ "transformer_blocks.14.attn2.to_k.weight",
553
+ "transformer_blocks.14.attn2.to_out.0.weight",
554
+ "transformer_blocks.14.attn2.to_q.weight",
555
+ "transformer_blocks.14.attn2.to_v.weight",
556
+ "transformer_blocks.14.audio_attn1.to_k.weight",
557
+ "transformer_blocks.14.audio_attn1.to_out.0.weight",
558
+ "transformer_blocks.14.audio_attn1.to_q.weight",
559
+ "transformer_blocks.14.audio_attn1.to_v.weight",
560
+ "transformer_blocks.14.audio_attn2.to_k.weight",
561
+ "transformer_blocks.14.audio_attn2.to_out.0.weight",
562
+ "transformer_blocks.14.audio_attn2.to_q.weight",
563
+ "transformer_blocks.14.audio_attn2.to_v.weight",
564
+ "transformer_blocks.14.audio_ff.net.0.proj.weight",
565
+ "transformer_blocks.14.audio_to_video_attn.to_k.weight",
566
+ "transformer_blocks.14.audio_to_video_attn.to_out.0.weight",
567
+ "transformer_blocks.14.audio_to_video_attn.to_q.weight",
568
+ "transformer_blocks.14.audio_to_video_attn.to_v.weight",
569
+ "transformer_blocks.14.ff.net.0.proj.weight",
570
+ "transformer_blocks.14.video_to_audio_attn.to_k.weight",
571
+ "transformer_blocks.14.video_to_audio_attn.to_out.0.weight",
572
+ "transformer_blocks.14.video_to_audio_attn.to_q.weight",
573
+ "transformer_blocks.14.video_to_audio_attn.to_v.weight",
574
+ "transformer_blocks.15.attn1.to_k.weight",
575
+ "transformer_blocks.15.attn1.to_out.0.weight",
576
+ "transformer_blocks.15.attn1.to_q.weight",
577
+ "transformer_blocks.15.attn1.to_v.weight",
578
+ "transformer_blocks.15.attn2.to_k.weight",
579
+ "transformer_blocks.15.attn2.to_out.0.weight",
580
+ "transformer_blocks.15.attn2.to_q.weight",
581
+ "transformer_blocks.15.attn2.to_v.weight",
582
+ "transformer_blocks.15.audio_attn1.to_k.weight",
583
+ "transformer_blocks.15.audio_attn1.to_out.0.weight",
584
+ "transformer_blocks.15.audio_attn1.to_q.weight",
585
+ "transformer_blocks.15.audio_attn1.to_v.weight",
586
+ "transformer_blocks.15.audio_attn2.to_k.weight",
587
+ "transformer_blocks.15.audio_attn2.to_out.0.weight",
588
+ "transformer_blocks.15.audio_attn2.to_q.weight",
589
+ "transformer_blocks.15.audio_attn2.to_v.weight",
590
+ "transformer_blocks.15.audio_ff.net.0.proj.weight",
591
+ "transformer_blocks.15.audio_to_video_attn.to_k.weight",
592
+ "transformer_blocks.15.audio_to_video_attn.to_out.0.weight",
593
+ "transformer_blocks.15.audio_to_video_attn.to_q.weight",
594
+ "transformer_blocks.15.audio_to_video_attn.to_v.weight",
595
+ "transformer_blocks.15.ff.net.0.proj.weight",
596
+ "transformer_blocks.15.video_to_audio_attn.to_k.weight",
597
+ "transformer_blocks.15.video_to_audio_attn.to_out.0.weight",
598
+ "transformer_blocks.15.video_to_audio_attn.to_q.weight",
599
+ "transformer_blocks.15.video_to_audio_attn.to_v.weight",
600
+ "transformer_blocks.16.attn1.to_k.weight",
601
+ "transformer_blocks.16.attn1.to_out.0.weight",
602
+ "transformer_blocks.16.attn1.to_q.weight",
603
+ "transformer_blocks.16.attn1.to_v.weight",
604
+ "transformer_blocks.16.attn2.to_k.weight",
605
+ "transformer_blocks.16.attn2.to_out.0.weight",
606
+ "transformer_blocks.16.attn2.to_q.weight",
607
+ "transformer_blocks.16.attn2.to_v.weight",
608
+ "transformer_blocks.16.audio_attn1.to_k.weight",
609
+ "transformer_blocks.16.audio_attn1.to_out.0.weight",
610
+ "transformer_blocks.16.audio_attn1.to_q.weight",
611
+ "transformer_blocks.16.audio_attn1.to_v.weight",
612
+ "transformer_blocks.16.audio_attn2.to_k.weight",
613
+ "transformer_blocks.16.audio_attn2.to_out.0.weight",
614
+ "transformer_blocks.16.audio_attn2.to_q.weight",
615
+ "transformer_blocks.16.audio_attn2.to_v.weight",
616
+ "transformer_blocks.16.audio_ff.net.0.proj.weight",
617
+ "transformer_blocks.16.audio_to_video_attn.to_k.weight",
618
+ "transformer_blocks.16.audio_to_video_attn.to_out.0.weight",
619
+ "transformer_blocks.16.audio_to_video_attn.to_q.weight",
620
+ "transformer_blocks.16.audio_to_video_attn.to_v.weight",
621
+ "transformer_blocks.16.ff.net.0.proj.weight",
622
+ "transformer_blocks.16.video_to_audio_attn.to_k.weight",
623
+ "transformer_blocks.16.video_to_audio_attn.to_out.0.weight",
624
+ "transformer_blocks.16.video_to_audio_attn.to_q.weight",
625
+ "transformer_blocks.16.video_to_audio_attn.to_v.weight",
626
+ "transformer_blocks.17.attn1.to_k.weight",
627
+ "transformer_blocks.17.attn1.to_out.0.weight",
628
+ "transformer_blocks.17.attn1.to_q.weight",
629
+ "transformer_blocks.17.attn1.to_v.weight",
630
+ "transformer_blocks.17.attn2.to_k.weight",
631
+ "transformer_blocks.17.attn2.to_out.0.weight",
632
+ "transformer_blocks.17.attn2.to_q.weight",
633
+ "transformer_blocks.17.attn2.to_v.weight",
634
+ "transformer_blocks.17.audio_attn1.to_k.weight",
635
+ "transformer_blocks.17.audio_attn1.to_out.0.weight",
636
+ "transformer_blocks.17.audio_attn1.to_q.weight",
637
+ "transformer_blocks.17.audio_attn1.to_v.weight",
638
+ "transformer_blocks.17.audio_attn2.to_k.weight",
639
+ "transformer_blocks.17.audio_attn2.to_out.0.weight",
640
+ "transformer_blocks.17.audio_attn2.to_q.weight",
641
+ "transformer_blocks.17.audio_attn2.to_v.weight",
642
+ "transformer_blocks.17.audio_ff.net.0.proj.weight",
643
+ "transformer_blocks.17.audio_to_video_attn.to_k.weight",
644
+ "transformer_blocks.17.audio_to_video_attn.to_out.0.weight",
645
+ "transformer_blocks.17.audio_to_video_attn.to_q.weight",
646
+ "transformer_blocks.17.audio_to_video_attn.to_v.weight",
647
+ "transformer_blocks.17.ff.net.0.proj.weight",
648
+ "transformer_blocks.17.video_to_audio_attn.to_k.weight",
649
+ "transformer_blocks.17.video_to_audio_attn.to_out.0.weight",
650
+ "transformer_blocks.17.video_to_audio_attn.to_q.weight",
651
+ "transformer_blocks.17.video_to_audio_attn.to_v.weight",
652
+ "transformer_blocks.18.attn1.to_k.weight",
653
+ "transformer_blocks.18.attn1.to_out.0.weight",
654
+ "transformer_blocks.18.attn1.to_q.weight",
655
+ "transformer_blocks.18.attn1.to_v.weight",
656
+ "transformer_blocks.18.attn2.to_out.0.weight",
657
+ "transformer_blocks.18.attn2.to_q.weight",
658
+ "transformer_blocks.18.attn2.to_v.weight",
659
+ "transformer_blocks.18.audio_attn1.to_k.weight",
660
+ "transformer_blocks.18.audio_attn1.to_out.0.weight",
661
+ "transformer_blocks.18.audio_attn1.to_q.weight",
662
+ "transformer_blocks.18.audio_attn1.to_v.weight",
663
+ "transformer_blocks.18.audio_attn2.to_k.weight",
664
+ "transformer_blocks.18.audio_attn2.to_out.0.weight",
665
+ "transformer_blocks.18.audio_attn2.to_q.weight",
666
+ "transformer_blocks.18.audio_attn2.to_v.weight",
667
+ "transformer_blocks.18.audio_to_video_attn.to_k.weight",
668
+ "transformer_blocks.18.audio_to_video_attn.to_q.weight",
669
+ "transformer_blocks.18.audio_ff.net.0.proj.weight",
670
+ "transformer_blocks.18.audio_ff.net.2.weight",
671
+ "transformer_blocks.18.audio_to_video_attn.to_out.0.weight",
672
+ "transformer_blocks.18.audio_to_video_attn.to_v.weight",
673
+ "transformer_blocks.18.ff.net.0.proj.weight",
674
+ "transformer_blocks.18.video_to_audio_attn.to_k.weight",
675
+ "transformer_blocks.18.video_to_audio_attn.to_out.0.weight",
676
+ "transformer_blocks.18.video_to_audio_attn.to_q.weight",
677
+ "transformer_blocks.18.video_to_audio_attn.to_v.weight",
678
+ "transformer_blocks.19.attn1.to_k.weight",
679
+ "transformer_blocks.19.attn1.to_out.0.weight",
680
+ "transformer_blocks.19.attn1.to_q.weight",
681
+ "transformer_blocks.19.attn1.to_v.weight",
682
+ "transformer_blocks.19.attn2.to_out.0.weight",
683
+ "transformer_blocks.19.attn2.to_q.weight",
684
+ "transformer_blocks.19.attn2.to_v.weight",
685
+ "transformer_blocks.19.audio_attn1.to_k.weight",
686
+ "transformer_blocks.19.audio_attn1.to_out.0.weight",
687
+ "transformer_blocks.19.audio_attn1.to_q.weight",
688
+ "transformer_blocks.19.audio_attn1.to_v.weight",
689
+ "transformer_blocks.19.audio_attn2.to_k.weight",
690
+ "transformer_blocks.19.audio_attn2.to_out.0.weight",
691
+ "transformer_blocks.19.audio_attn2.to_q.weight",
692
+ "transformer_blocks.19.audio_attn2.to_v.weight",
693
+ "transformer_blocks.19.audio_ff.net.0.proj.weight",
694
+ "transformer_blocks.19.audio_ff.net.2.weight",
695
+ "transformer_blocks.19.audio_to_video_attn.to_k.weight",
696
+ "transformer_blocks.19.audio_to_video_attn.to_out.0.weight",
697
+ "transformer_blocks.19.audio_to_video_attn.to_q.weight",
698
+ "transformer_blocks.19.audio_to_video_attn.to_v.weight",
699
+ "transformer_blocks.19.ff.net.0.proj.weight",
700
+ "transformer_blocks.19.video_to_audio_attn.to_k.weight",
701
+ "transformer_blocks.19.video_to_audio_attn.to_out.0.weight",
702
+ "transformer_blocks.19.video_to_audio_attn.to_q.weight",
703
+ "transformer_blocks.19.video_to_audio_attn.to_v.weight",
704
+ "transformer_blocks.20.attn1.to_k.weight",
705
+ "transformer_blocks.20.attn1.to_out.0.weight",
706
+ "transformer_blocks.20.attn1.to_q.weight",
707
+ "transformer_blocks.20.attn1.to_v.weight",
708
+ "transformer_blocks.20.attn2.to_out.0.weight",
709
+ "transformer_blocks.20.attn2.to_q.weight",
710
+ "transformer_blocks.20.attn2.to_v.weight",
711
+ "transformer_blocks.20.audio_attn1.to_k.weight",
712
+ "transformer_blocks.20.audio_attn1.to_out.0.weight",
713
+ "transformer_blocks.20.audio_attn1.to_q.weight",
714
+ "transformer_blocks.20.audio_attn1.to_v.weight",
715
+ "transformer_blocks.20.audio_attn2.to_k.weight",
716
+ "transformer_blocks.20.audio_attn2.to_out.0.weight",
717
+ "transformer_blocks.20.audio_attn2.to_q.weight",
718
+ "transformer_blocks.20.audio_attn2.to_v.weight",
719
+ "transformer_blocks.20.audio_ff.net.0.proj.weight",
720
+ "transformer_blocks.20.audio_ff.net.2.weight",
721
+ "transformer_blocks.20.audio_to_video_attn.to_k.weight",
722
+ "transformer_blocks.20.audio_to_video_attn.to_out.0.weight",
723
+ "transformer_blocks.20.audio_to_video_attn.to_q.weight",
724
+ "transformer_blocks.20.audio_to_video_attn.to_v.weight",
725
+ "transformer_blocks.20.ff.net.0.proj.weight",
726
+ "transformer_blocks.20.video_to_audio_attn.to_k.weight",
727
+ "transformer_blocks.20.video_to_audio_attn.to_out.0.weight",
728
+ "transformer_blocks.20.video_to_audio_attn.to_q.weight",
729
+ "transformer_blocks.20.video_to_audio_attn.to_v.weight",
730
+ "transformer_blocks.21.attn1.to_k.weight",
731
+ "transformer_blocks.21.attn1.to_out.0.weight",
732
+ "transformer_blocks.21.attn1.to_q.weight",
733
+ "transformer_blocks.21.attn1.to_v.weight",
734
+ "transformer_blocks.21.attn2.to_out.0.weight",
735
+ "transformer_blocks.21.attn2.to_q.weight",
736
+ "transformer_blocks.21.attn2.to_v.weight",
737
+ "transformer_blocks.21.audio_attn1.to_k.weight",
738
+ "transformer_blocks.21.audio_attn1.to_out.0.weight",
739
+ "transformer_blocks.21.audio_attn1.to_q.weight",
740
+ "transformer_blocks.21.audio_attn1.to_v.weight",
741
+ "transformer_blocks.21.audio_attn2.to_k.weight",
742
+ "transformer_blocks.21.audio_attn2.to_out.0.weight",
743
+ "transformer_blocks.21.audio_attn2.to_q.weight",
744
+ "transformer_blocks.21.audio_attn2.to_v.weight",
745
+ "transformer_blocks.21.audio_ff.net.0.proj.weight",
746
+ "transformer_blocks.21.audio_ff.net.2.weight",
747
+ "transformer_blocks.21.audio_to_video_attn.to_k.weight",
748
+ "transformer_blocks.21.audio_to_video_attn.to_out.0.weight",
749
+ "transformer_blocks.21.audio_to_video_attn.to_q.weight",
750
+ "transformer_blocks.21.audio_to_video_attn.to_v.weight",
751
+ "transformer_blocks.21.ff.net.0.proj.weight",
752
+ "transformer_blocks.21.ff.net.2.weight",
753
+ "transformer_blocks.21.video_to_audio_attn.to_k.weight",
754
+ "transformer_blocks.21.video_to_audio_attn.to_out.0.weight",
755
+ "transformer_blocks.21.video_to_audio_attn.to_q.weight",
756
+ "transformer_blocks.21.video_to_audio_attn.to_v.weight",
757
+ "transformer_blocks.22.attn1.to_k.weight",
758
+ "transformer_blocks.22.attn1.to_out.0.weight",
759
+ "transformer_blocks.22.attn1.to_q.weight",
760
+ "transformer_blocks.22.attn1.to_v.weight",
761
+ "transformer_blocks.22.attn2.to_out.0.weight",
762
+ "transformer_blocks.22.attn2.to_q.weight",
763
+ "transformer_blocks.22.attn2.to_v.weight",
764
+ "transformer_blocks.22.audio_attn1.to_k.weight",
765
+ "transformer_blocks.22.audio_attn1.to_out.0.weight",
766
+ "transformer_blocks.22.audio_attn1.to_q.weight",
767
+ "transformer_blocks.22.audio_attn1.to_v.weight",
768
+ "transformer_blocks.22.audio_attn2.to_k.weight",
769
+ "transformer_blocks.22.audio_attn2.to_out.0.weight",
770
+ "transformer_blocks.22.audio_attn2.to_q.weight",
771
+ "transformer_blocks.22.audio_attn2.to_v.weight",
772
+ "transformer_blocks.22.audio_ff.net.0.proj.weight",
773
+ "transformer_blocks.22.audio_ff.net.2.weight",
774
+ "transformer_blocks.22.audio_to_video_attn.to_k.weight",
775
+ "transformer_blocks.22.audio_to_video_attn.to_out.0.weight",
776
+ "transformer_blocks.22.audio_to_video_attn.to_q.weight",
777
+ "transformer_blocks.22.audio_to_video_attn.to_v.weight",
778
+ "transformer_blocks.22.ff.net.0.proj.weight",
779
+ "transformer_blocks.22.ff.net.2.weight",
780
+ "transformer_blocks.22.video_to_audio_attn.to_k.weight",
781
+ "transformer_blocks.22.video_to_audio_attn.to_out.0.weight",
782
+ "transformer_blocks.22.video_to_audio_attn.to_q.weight",
783
+ "transformer_blocks.22.video_to_audio_attn.to_v.weight",
784
+ "transformer_blocks.23.attn1.to_k.weight",
785
+ "transformer_blocks.23.attn1.to_out.0.weight",
786
+ "transformer_blocks.23.attn1.to_q.weight",
787
+ "transformer_blocks.23.attn1.to_v.weight",
788
+ "transformer_blocks.23.attn2.to_out.0.weight",
789
+ "transformer_blocks.23.attn2.to_q.weight",
790
+ "transformer_blocks.23.audio_attn1.to_k.weight",
791
+ "transformer_blocks.23.audio_attn1.to_out.0.weight",
792
+ "transformer_blocks.23.audio_attn1.to_q.weight",
793
+ "transformer_blocks.23.audio_attn1.to_v.weight",
794
+ "transformer_blocks.23.audio_attn2.to_k.weight",
795
+ "transformer_blocks.23.audio_attn2.to_out.0.weight",
796
+ "transformer_blocks.23.audio_attn2.to_q.weight",
797
+ "transformer_blocks.23.audio_attn2.to_v.weight",
798
+ "transformer_blocks.23.audio_ff.net.0.proj.weight",
799
+ "transformer_blocks.23.audio_ff.net.2.weight",
800
+ "transformer_blocks.23.audio_to_video_attn.to_k.weight",
801
+ "transformer_blocks.23.audio_to_video_attn.to_out.0.weight",
802
+ "transformer_blocks.23.audio_to_video_attn.to_q.weight",
803
+ "transformer_blocks.23.audio_to_video_attn.to_v.weight",
804
+ "transformer_blocks.23.ff.net.0.proj.weight",
805
+ "transformer_blocks.23.ff.net.2.weight",
806
+ "transformer_blocks.23.video_to_audio_attn.to_k.weight",
807
+ "transformer_blocks.23.video_to_audio_attn.to_out.0.weight",
808
+ "transformer_blocks.23.video_to_audio_attn.to_q.weight",
809
+ "transformer_blocks.23.video_to_audio_attn.to_v.weight",
810
+ "transformer_blocks.24.attn1.to_k.weight",
811
+ "transformer_blocks.24.attn1.to_out.0.weight",
812
+ "transformer_blocks.24.attn1.to_q.weight",
813
+ "transformer_blocks.24.attn1.to_v.weight",
814
+ "transformer_blocks.24.attn2.to_out.0.weight",
815
+ "transformer_blocks.24.attn2.to_q.weight",
816
+ "transformer_blocks.24.audio_attn1.to_k.weight",
817
+ "transformer_blocks.24.audio_attn1.to_out.0.weight",
818
+ "transformer_blocks.24.audio_attn1.to_q.weight",
819
+ "transformer_blocks.24.audio_attn1.to_v.weight",
820
+ "transformer_blocks.24.audio_attn2.to_k.weight",
821
+ "transformer_blocks.24.audio_attn2.to_out.0.weight",
822
+ "transformer_blocks.24.audio_attn2.to_q.weight",
823
+ "transformer_blocks.24.audio_attn2.to_v.weight",
824
+ "transformer_blocks.24.audio_to_video_attn.to_k.weight",
825
+ "transformer_blocks.24.audio_to_video_attn.to_out.0.weight",
826
+ "transformer_blocks.24.audio_to_video_attn.to_q.weight",
827
+ "transformer_blocks.24.audio_to_video_attn.to_v.weight",
828
+ "transformer_blocks.24.ff.net.0.proj.weight",
829
+ "transformer_blocks.24.ff.net.2.weight",
830
+ "transformer_blocks.24.video_to_audio_attn.to_k.weight",
831
+ "transformer_blocks.24.video_to_audio_attn.to_out.0.weight",
832
+ "transformer_blocks.24.video_to_audio_attn.to_q.weight",
833
+ "transformer_blocks.24.video_to_audio_attn.to_v.weight",
834
+ "transformer_blocks.24.audio_ff.net.0.proj.weight",
835
+ "transformer_blocks.24.audio_ff.net.2.weight",
836
+ "transformer_blocks.25.attn1.to_k.weight",
837
+ "transformer_blocks.25.attn1.to_out.0.weight",
838
+ "transformer_blocks.25.attn1.to_q.weight",
839
+ "transformer_blocks.25.attn1.to_v.weight",
840
+ "transformer_blocks.25.attn2.to_out.0.weight",
841
+ "transformer_blocks.25.attn2.to_q.weight",
842
+ "transformer_blocks.25.audio_attn1.to_k.weight",
843
+ "transformer_blocks.25.audio_attn1.to_out.0.weight",
844
+ "transformer_blocks.25.audio_attn1.to_q.weight",
845
+ "transformer_blocks.25.audio_attn2.to_k.weight",
846
+ "transformer_blocks.25.audio_attn2.to_out.0.weight",
847
+ "transformer_blocks.25.audio_attn2.to_q.weight",
848
+ "transformer_blocks.25.audio_attn2.to_v.weight",
849
+ "transformer_blocks.25.audio_ff.net.0.proj.weight",
850
+ "transformer_blocks.25.audio_ff.net.2.weight",
851
+ "transformer_blocks.25.audio_to_video_attn.to_k.weight",
852
+ "transformer_blocks.25.audio_to_video_attn.to_out.0.weight",
853
+ "transformer_blocks.25.audio_to_video_attn.to_q.weight",
854
+ "transformer_blocks.25.ff.net.0.proj.weight",
855
+ "transformer_blocks.25.ff.net.2.weight",
856
+ "transformer_blocks.25.video_to_audio_attn.to_k.weight",
857
+ "transformer_blocks.25.video_to_audio_attn.to_out.0.weight",
858
+ "transformer_blocks.25.video_to_audio_attn.to_q.weight",
859
+ "transformer_blocks.25.video_to_audio_attn.to_v.weight",
860
+ "transformer_blocks.26.attn1.to_k.weight",
861
+ "transformer_blocks.26.attn1.to_out.0.weight",
862
+ "transformer_blocks.26.attn1.to_q.weight",
863
+ "transformer_blocks.26.attn1.to_v.weight",
864
+ "transformer_blocks.26.attn2.to_out.0.weight",
865
+ "transformer_blocks.26.attn2.to_q.weight",
866
+ "transformer_blocks.26.audio_attn1.to_k.weight",
867
+ "transformer_blocks.26.audio_attn1.to_out.0.weight",
868
+ "transformer_blocks.26.audio_attn1.to_q.weight",
869
+ "transformer_blocks.26.audio_attn1.to_v.weight",
870
+ "transformer_blocks.26.audio_attn2.to_k.weight",
871
+ "transformer_blocks.26.audio_attn2.to_out.0.weight",
872
+ "transformer_blocks.26.audio_attn2.to_q.weight",
873
+ "transformer_blocks.26.audio_attn2.to_v.weight",
874
+ "transformer_blocks.26.audio_ff.net.0.proj.weight",
875
+ "transformer_blocks.26.audio_ff.net.2.weight",
876
+ "transformer_blocks.26.audio_to_video_attn.to_k.weight",
877
+ "transformer_blocks.26.audio_to_video_attn.to_out.0.weight",
878
+ "transformer_blocks.26.audio_to_video_attn.to_q.weight",
879
+ "transformer_blocks.26.audio_to_video_attn.to_v.weight",
880
+ "transformer_blocks.26.ff.net.0.proj.weight",
881
+ "transformer_blocks.26.ff.net.2.weight",
882
+ "transformer_blocks.26.video_to_audio_attn.to_k.weight",
883
+ "transformer_blocks.26.video_to_audio_attn.to_out.0.weight",
884
+ "transformer_blocks.26.video_to_audio_attn.to_q.weight",
885
+ "transformer_blocks.26.video_to_audio_attn.to_v.weight",
886
+ "transformer_blocks.27.attn1.to_k.weight",
887
+ "transformer_blocks.27.attn1.to_out.0.weight",
888
+ "transformer_blocks.27.attn1.to_q.weight",
889
+ "transformer_blocks.27.attn1.to_v.weight",
890
+ "transformer_blocks.27.attn2.to_out.0.weight",
891
+ "transformer_blocks.27.attn2.to_q.weight",
892
+ "transformer_blocks.27.audio_attn1.to_k.weight",
893
+ "transformer_blocks.27.audio_attn1.to_out.0.weight",
894
+ "transformer_blocks.27.audio_attn1.to_q.weight",
895
+ "transformer_blocks.27.audio_attn2.to_k.weight",
896
+ "transformer_blocks.27.audio_attn2.to_out.0.weight",
897
+ "transformer_blocks.27.audio_attn2.to_q.weight",
898
+ "transformer_blocks.27.audio_attn2.to_v.weight",
899
+ "transformer_blocks.27.audio_ff.net.0.proj.weight",
900
+ "transformer_blocks.27.audio_ff.net.2.weight",
901
+ "transformer_blocks.27.audio_to_video_attn.to_k.weight",
902
+ "transformer_blocks.27.audio_to_video_attn.to_out.0.weight",
903
+ "transformer_blocks.27.audio_to_video_attn.to_q.weight",
904
+ "transformer_blocks.27.ff.net.0.proj.weight",
905
+ "transformer_blocks.27.ff.net.2.weight",
906
+ "transformer_blocks.27.video_to_audio_attn.to_k.weight",
907
+ "transformer_blocks.27.video_to_audio_attn.to_out.0.weight",
908
+ "transformer_blocks.27.video_to_audio_attn.to_q.weight",
909
+ "transformer_blocks.27.video_to_audio_attn.to_v.weight",
910
+ "transformer_blocks.28.attn1.to_k.weight",
911
+ "transformer_blocks.28.attn1.to_out.0.weight",
912
+ "transformer_blocks.28.attn1.to_q.weight",
913
+ "transformer_blocks.28.attn1.to_v.weight",
914
+ "transformer_blocks.28.attn2.to_out.0.weight",
915
+ "transformer_blocks.28.attn2.to_q.weight",
916
+ "transformer_blocks.28.attn2.to_v.weight",
917
+ "transformer_blocks.28.audio_attn1.to_k.weight",
918
+ "transformer_blocks.28.audio_attn1.to_out.0.weight",
919
+ "transformer_blocks.28.audio_attn1.to_q.weight",
920
+ "transformer_blocks.28.audio_attn2.to_k.weight",
921
+ "transformer_blocks.28.audio_attn2.to_out.0.weight",
922
+ "transformer_blocks.28.audio_attn2.to_q.weight",
923
+ "transformer_blocks.28.audio_attn2.to_v.weight",
924
+ "transformer_blocks.28.audio_ff.net.0.proj.weight",
925
+ "transformer_blocks.28.audio_ff.net.2.weight",
926
+ "transformer_blocks.28.audio_to_video_attn.to_k.weight",
927
+ "transformer_blocks.28.audio_to_video_attn.to_out.0.weight",
928
+ "transformer_blocks.28.audio_to_video_attn.to_q.weight",
929
+ "transformer_blocks.28.audio_to_video_attn.to_v.weight",
930
+ "transformer_blocks.28.ff.net.0.proj.weight",
931
+ "transformer_blocks.28.ff.net.2.weight",
932
+ "transformer_blocks.28.video_to_audio_attn.to_k.weight",
933
+ "transformer_blocks.28.video_to_audio_attn.to_out.0.weight",
934
+ "transformer_blocks.28.video_to_audio_attn.to_q.weight",
935
+ "transformer_blocks.28.video_to_audio_attn.to_v.weight",
936
+ "transformer_blocks.29.attn1.to_out.0.weight",
937
+ "transformer_blocks.29.attn1.to_q.weight",
938
+ "transformer_blocks.29.attn1.to_v.weight",
939
+ "transformer_blocks.29.attn2.to_out.0.weight",
940
+ "transformer_blocks.29.audio_attn1.to_k.weight",
941
+ "transformer_blocks.29.audio_attn1.to_out.0.weight",
942
+ "transformer_blocks.29.audio_attn1.to_q.weight",
943
+ "transformer_blocks.29.audio_attn1.to_v.weight",
944
+ "transformer_blocks.29.audio_attn2.to_k.weight",
945
+ "transformer_blocks.29.audio_attn2.to_out.0.weight",
946
+ "transformer_blocks.29.audio_attn2.to_q.weight",
947
+ "transformer_blocks.29.audio_attn2.to_v.weight",
948
+ "transformer_blocks.29.audio_ff.net.0.proj.weight",
949
+ "transformer_blocks.29.audio_ff.net.2.weight",
950
+ "transformer_blocks.29.audio_to_video_attn.to_k.weight",
951
+ "transformer_blocks.29.audio_to_video_attn.to_out.0.weight",
952
+ "transformer_blocks.29.audio_to_video_attn.to_q.weight",
953
+ "transformer_blocks.29.audio_to_video_attn.to_v.weight",
954
+ "transformer_blocks.29.ff.net.0.proj.weight",
955
+ "transformer_blocks.29.ff.net.2.weight",
956
+ "transformer_blocks.29.video_to_audio_attn.to_k.weight",
957
+ "transformer_blocks.29.video_to_audio_attn.to_out.0.weight",
958
+ "transformer_blocks.29.video_to_audio_attn.to_q.weight",
959
+ "transformer_blocks.30.attn1.to_out.0.weight",
960
+ "transformer_blocks.30.attn1.to_q.weight",
961
+ "transformer_blocks.30.attn1.to_v.weight",
962
+ "transformer_blocks.30.attn2.to_out.0.weight",
963
+ "transformer_blocks.30.attn2.to_q.weight",
964
+ "transformer_blocks.30.audio_attn1.to_k.weight",
965
+ "transformer_blocks.30.audio_attn1.to_out.0.weight",
966
+ "transformer_blocks.30.audio_attn1.to_q.weight",
967
+ "transformer_blocks.30.audio_attn1.to_v.weight",
968
+ "transformer_blocks.30.audio_attn2.to_k.weight",
969
+ "transformer_blocks.30.audio_attn2.to_out.0.weight",
970
+ "transformer_blocks.30.audio_attn2.to_q.weight",
971
+ "transformer_blocks.30.audio_attn2.to_v.weight",
972
+ "transformer_blocks.30.audio_ff.net.0.proj.weight",
973
+ "transformer_blocks.30.audio_ff.net.2.weight",
974
+ "transformer_blocks.30.audio_to_video_attn.to_k.weight",
975
+ "transformer_blocks.30.audio_to_video_attn.to_out.0.weight",
976
+ "transformer_blocks.30.audio_to_video_attn.to_q.weight",
977
+ "transformer_blocks.30.audio_to_video_attn.to_v.weight",
978
+ "transformer_blocks.30.ff.net.0.proj.weight",
979
+ "transformer_blocks.30.ff.net.2.weight",
980
+ "transformer_blocks.30.video_to_audio_attn.to_k.weight",
981
+ "transformer_blocks.30.video_to_audio_attn.to_out.0.weight",
982
+ "transformer_blocks.30.video_to_audio_attn.to_q.weight",
983
+ "transformer_blocks.30.video_to_audio_attn.to_v.weight",
984
+ "transformer_blocks.31.attn1.to_out.0.weight",
985
+ "transformer_blocks.31.attn1.to_q.weight",
986
+ "transformer_blocks.31.attn1.to_v.weight",
987
+ "transformer_blocks.31.attn2.to_q.weight",
988
+ "transformer_blocks.31.audio_attn1.to_k.weight",
989
+ "transformer_blocks.31.audio_attn1.to_out.0.weight",
990
+ "transformer_blocks.31.audio_attn1.to_q.weight",
991
+ "transformer_blocks.31.audio_attn1.to_v.weight",
992
+ "transformer_blocks.31.attn2.to_out.0.weight",
993
+ "transformer_blocks.31.audio_attn2.to_k.weight",
994
+ "transformer_blocks.31.audio_attn2.to_out.0.weight",
995
+ "transformer_blocks.31.audio_attn2.to_q.weight",
996
+ "transformer_blocks.31.audio_attn2.to_v.weight",
997
+ "transformer_blocks.31.audio_ff.net.0.proj.weight",
998
+ "transformer_blocks.31.audio_ff.net.2.weight",
999
+ "transformer_blocks.31.audio_to_video_attn.to_k.weight",
1000
+ "transformer_blocks.31.audio_to_video_attn.to_out.0.weight",
1001
+ "transformer_blocks.31.audio_to_video_attn.to_q.weight",
1002
+ "transformer_blocks.31.audio_to_video_attn.to_v.weight",
1003
+ "transformer_blocks.31.ff.net.0.proj.weight",
1004
+ "transformer_blocks.31.ff.net.2.weight",
1005
+ "transformer_blocks.31.video_to_audio_attn.to_k.weight",
1006
+ "transformer_blocks.31.video_to_audio_attn.to_out.0.weight",
1007
+ "transformer_blocks.31.video_to_audio_attn.to_q.weight",
1008
+ "transformer_blocks.31.video_to_audio_attn.to_v.weight",
1009
+ "transformer_blocks.32.attn1.to_k.weight",
1010
+ "transformer_blocks.32.attn1.to_out.0.weight",
1011
+ "transformer_blocks.32.attn1.to_q.weight",
1012
+ "transformer_blocks.32.attn1.to_v.weight",
1013
+ "transformer_blocks.32.attn2.to_out.0.weight",
1014
+ "transformer_blocks.32.attn2.to_q.weight",
1015
+ "transformer_blocks.32.audio_attn1.to_k.weight",
1016
+ "transformer_blocks.32.audio_attn1.to_out.0.weight",
1017
+ "transformer_blocks.32.audio_attn1.to_q.weight",
1018
+ "transformer_blocks.32.audio_attn1.to_v.weight",
1019
+ "transformer_blocks.32.audio_attn2.to_k.weight",
1020
+ "transformer_blocks.32.audio_attn2.to_out.0.weight",
1021
+ "transformer_blocks.32.audio_attn2.to_q.weight",
1022
+ "transformer_blocks.32.audio_attn2.to_v.weight",
1023
+ "transformer_blocks.32.audio_ff.net.0.proj.weight",
1024
+ "transformer_blocks.32.audio_ff.net.2.weight",
1025
+ "transformer_blocks.32.audio_to_video_attn.to_k.weight",
1026
+ "transformer_blocks.32.audio_to_video_attn.to_out.0.weight",
1027
+ "transformer_blocks.32.audio_to_video_attn.to_q.weight",
1028
+ "transformer_blocks.32.audio_to_video_attn.to_v.weight",
1029
+ "transformer_blocks.32.ff.net.0.proj.weight",
1030
+ "transformer_blocks.32.ff.net.2.weight",
1031
+ "transformer_blocks.32.video_to_audio_attn.to_k.weight",
1032
+ "transformer_blocks.32.video_to_audio_attn.to_out.0.weight",
1033
+ "transformer_blocks.32.video_to_audio_attn.to_q.weight",
1034
+ "transformer_blocks.33.attn1.to_out.0.weight",
1035
+ "transformer_blocks.33.attn1.to_q.weight",
1036
+ "transformer_blocks.33.attn1.to_v.weight",
1037
+ "transformer_blocks.33.attn2.to_out.0.weight",
1038
+ "transformer_blocks.33.attn2.to_q.weight",
1039
+ "transformer_blocks.33.audio_attn1.to_k.weight",
1040
+ "transformer_blocks.33.audio_attn1.to_out.0.weight",
1041
+ "transformer_blocks.33.audio_attn1.to_q.weight",
1042
+ "transformer_blocks.33.audio_attn1.to_v.weight",
1043
+ "transformer_blocks.33.audio_attn2.to_k.weight",
1044
+ "transformer_blocks.33.audio_attn2.to_out.0.weight",
1045
+ "transformer_blocks.33.audio_attn2.to_q.weight",
1046
+ "transformer_blocks.33.audio_attn2.to_v.weight",
1047
+ "transformer_blocks.33.audio_ff.net.0.proj.weight",
1048
+ "transformer_blocks.33.audio_ff.net.2.weight",
1049
+ "transformer_blocks.33.audio_to_video_attn.to_k.weight",
1050
+ "transformer_blocks.33.audio_to_video_attn.to_out.0.weight",
1051
+ "transformer_blocks.33.audio_to_video_attn.to_q.weight",
1052
+ "transformer_blocks.33.audio_to_video_attn.to_v.weight",
1053
+ "transformer_blocks.33.ff.net.0.proj.weight",
1054
+ "transformer_blocks.33.ff.net.2.weight",
1055
+ "transformer_blocks.33.video_to_audio_attn.to_k.weight",
1056
+ "transformer_blocks.33.video_to_audio_attn.to_out.0.weight",
1057
+ "transformer_blocks.33.video_to_audio_attn.to_q.weight",
1058
+ "transformer_blocks.33.video_to_audio_attn.to_v.weight",
1059
+ "transformer_blocks.34.attn1.to_k.weight",
1060
+ "transformer_blocks.34.attn1.to_out.0.weight",
1061
+ "transformer_blocks.34.attn1.to_q.weight",
1062
+ "transformer_blocks.34.attn1.to_v.weight",
1063
+ "transformer_blocks.34.attn2.to_q.weight",
1064
+ "transformer_blocks.34.audio_attn1.to_k.weight",
1065
+ "transformer_blocks.34.audio_attn1.to_out.0.weight",
1066
+ "transformer_blocks.34.audio_attn1.to_q.weight",
1067
+ "transformer_blocks.34.audio_attn1.to_v.weight",
1068
+ "transformer_blocks.34.audio_attn2.to_k.weight",
1069
+ "transformer_blocks.34.audio_attn2.to_out.0.weight",
1070
+ "transformer_blocks.34.audio_attn2.to_q.weight",
1071
+ "transformer_blocks.34.audio_attn2.to_v.weight",
1072
+ "transformer_blocks.34.audio_ff.net.0.proj.weight",
1073
+ "transformer_blocks.34.audio_ff.net.2.weight",
1074
+ "transformer_blocks.34.audio_to_video_attn.to_k.weight",
1075
+ "transformer_blocks.34.audio_to_video_attn.to_out.0.weight",
1076
+ "transformer_blocks.34.audio_to_video_attn.to_q.weight",
1077
+ "transformer_blocks.34.audio_to_video_attn.to_v.weight",
1078
+ "transformer_blocks.34.ff.net.0.proj.weight",
1079
+ "transformer_blocks.34.ff.net.2.weight",
1080
+ "transformer_blocks.34.video_to_audio_attn.to_k.weight",
1081
+ "transformer_blocks.34.video_to_audio_attn.to_out.0.weight",
1082
+ "transformer_blocks.34.video_to_audio_attn.to_q.weight",
1083
+ "transformer_blocks.34.video_to_audio_attn.to_v.weight",
1084
+ "transformer_blocks.35.attn1.to_k.weight",
1085
+ "transformer_blocks.35.attn1.to_out.0.weight",
1086
+ "transformer_blocks.35.attn1.to_q.weight",
1087
+ "transformer_blocks.35.attn1.to_v.weight",
1088
+ "transformer_blocks.35.attn2.to_q.weight",
1089
+ "transformer_blocks.35.audio_attn1.to_k.weight",
1090
+ "transformer_blocks.35.audio_attn1.to_out.0.weight",
1091
+ "transformer_blocks.35.audio_attn1.to_q.weight",
1092
+ "transformer_blocks.35.audio_attn1.to_v.weight",
1093
+ "transformer_blocks.35.audio_attn2.to_k.weight",
1094
+ "transformer_blocks.35.audio_attn2.to_out.0.weight",
1095
+ "transformer_blocks.35.audio_attn2.to_q.weight",
1096
+ "transformer_blocks.35.audio_attn2.to_v.weight",
1097
+ "transformer_blocks.35.audio_ff.net.0.proj.weight",
1098
+ "transformer_blocks.35.audio_ff.net.2.weight",
1099
+ "transformer_blocks.35.audio_to_video_attn.to_k.weight",
1100
+ "transformer_blocks.35.audio_to_video_attn.to_out.0.weight",
1101
+ "transformer_blocks.35.audio_to_video_attn.to_q.weight",
1102
+ "transformer_blocks.35.audio_to_video_attn.to_v.weight",
1103
+ "transformer_blocks.35.ff.net.0.proj.weight",
1104
+ "transformer_blocks.35.ff.net.2.weight",
1105
+ "transformer_blocks.35.video_to_audio_attn.to_k.weight",
1106
+ "transformer_blocks.35.video_to_audio_attn.to_out.0.weight",
1107
+ "transformer_blocks.35.video_to_audio_attn.to_q.weight",
1108
+ "transformer_blocks.35.video_to_audio_attn.to_v.weight",
1109
+ "transformer_blocks.36.attn1.to_k.weight",
1110
+ "transformer_blocks.36.attn1.to_out.0.weight",
1111
+ "transformer_blocks.36.attn1.to_q.weight",
1112
+ "transformer_blocks.36.attn1.to_v.weight",
1113
+ "transformer_blocks.36.attn2.to_q.weight",
1114
+ "transformer_blocks.36.audio_attn1.to_k.weight",
1115
+ "transformer_blocks.36.audio_attn1.to_out.0.weight",
1116
+ "transformer_blocks.36.audio_attn1.to_q.weight",
1117
+ "transformer_blocks.36.audio_attn2.to_k.weight",
1118
+ "transformer_blocks.36.audio_attn2.to_out.0.weight",
1119
+ "transformer_blocks.36.audio_attn2.to_q.weight",
1120
+ "transformer_blocks.36.audio_attn2.to_v.weight",
1121
+ "transformer_blocks.36.audio_ff.net.0.proj.weight",
1122
+ "transformer_blocks.36.audio_ff.net.2.weight",
1123
+ "transformer_blocks.36.audio_to_video_attn.to_k.weight",
1124
+ "transformer_blocks.36.audio_to_video_attn.to_out.0.weight",
1125
+ "transformer_blocks.36.audio_to_video_attn.to_q.weight",
1126
+ "transformer_blocks.36.ff.net.0.proj.weight",
1127
+ "transformer_blocks.36.ff.net.2.weight",
1128
+ "transformer_blocks.36.video_to_audio_attn.to_k.weight",
1129
+ "transformer_blocks.36.video_to_audio_attn.to_out.0.weight",
1130
+ "transformer_blocks.36.video_to_audio_attn.to_q.weight",
1131
+ "transformer_blocks.36.video_to_audio_attn.to_v.weight",
1132
+ "transformer_blocks.37.attn1.to_out.0.weight",
1133
+ "transformer_blocks.37.attn1.to_q.weight",
1134
+ "transformer_blocks.37.attn2.to_q.weight",
1135
+ "transformer_blocks.37.audio_attn1.to_k.weight",
1136
+ "transformer_blocks.37.audio_attn1.to_out.0.weight",
1137
+ "transformer_blocks.37.audio_attn1.to_q.weight",
1138
+ "transformer_blocks.37.audio_attn1.to_v.weight",
1139
+ "transformer_blocks.37.audio_attn2.to_k.weight",
1140
+ "transformer_blocks.37.audio_attn2.to_out.0.weight",
1141
+ "transformer_blocks.37.audio_attn2.to_q.weight",
1142
+ "transformer_blocks.37.audio_attn2.to_v.weight",
1143
+ "transformer_blocks.37.audio_to_video_attn.to_k.weight",
1144
+ "transformer_blocks.37.audio_to_video_attn.to_out.0.weight",
1145
+ "transformer_blocks.37.audio_to_video_attn.to_q.weight",
1146
+ "transformer_blocks.37.audio_to_video_attn.to_v.weight",
1147
+ "transformer_blocks.37.ff.net.0.proj.weight",
1148
+ "transformer_blocks.37.video_to_audio_attn.to_k.weight",
1149
+ "transformer_blocks.37.video_to_audio_attn.to_out.0.weight",
1150
+ "transformer_blocks.37.video_to_audio_attn.to_q.weight",
1151
+ "transformer_blocks.37.video_to_audio_attn.to_v.weight",
1152
+ "transformer_blocks.37.audio_ff.net.0.proj.weight",
1153
+ "transformer_blocks.37.audio_ff.net.2.weight",
1154
+ "transformer_blocks.37.ff.net.2.weight",
1155
+ "transformer_blocks.38.attn1.to_out.0.weight",
1156
+ "transformer_blocks.38.attn1.to_q.weight",
1157
+ "transformer_blocks.38.attn2.to_q.weight",
1158
+ "transformer_blocks.38.audio_attn1.to_k.weight",
1159
+ "transformer_blocks.38.audio_attn1.to_out.0.weight",
1160
+ "transformer_blocks.38.audio_attn1.to_q.weight",
1161
+ "transformer_blocks.38.audio_attn2.to_k.weight",
1162
+ "transformer_blocks.38.audio_attn2.to_out.0.weight",
1163
+ "transformer_blocks.38.audio_attn2.to_q.weight",
1164
+ "transformer_blocks.38.audio_attn2.to_v.weight",
1165
+ "transformer_blocks.38.audio_ff.net.0.proj.weight",
1166
+ "transformer_blocks.38.audio_ff.net.2.weight",
1167
+ "transformer_blocks.38.audio_to_video_attn.to_k.weight",
1168
+ "transformer_blocks.38.audio_to_video_attn.to_out.0.weight",
1169
+ "transformer_blocks.38.audio_to_video_attn.to_q.weight",
1170
+ "transformer_blocks.38.ff.net.0.proj.weight",
1171
+ "transformer_blocks.38.ff.net.2.weight",
1172
+ "transformer_blocks.38.video_to_audio_attn.to_k.weight",
1173
+ "transformer_blocks.38.video_to_audio_attn.to_out.0.weight",
1174
+ "transformer_blocks.38.video_to_audio_attn.to_q.weight",
1175
+ "transformer_blocks.38.video_to_audio_attn.to_v.weight",
1176
+ "transformer_blocks.39.attn1.to_out.0.weight",
1177
+ "transformer_blocks.39.attn1.to_q.weight",
1178
+ "transformer_blocks.39.attn2.to_q.weight",
1179
+ "transformer_blocks.39.audio_attn1.to_k.weight",
1180
+ "transformer_blocks.39.audio_attn1.to_out.0.weight",
1181
+ "transformer_blocks.39.audio_attn1.to_q.weight",
1182
+ "transformer_blocks.39.audio_attn1.to_v.weight",
1183
+ "transformer_blocks.39.audio_attn2.to_k.weight",
1184
+ "transformer_blocks.39.audio_attn2.to_out.0.weight",
1185
+ "transformer_blocks.39.audio_attn2.to_q.weight",
1186
+ "transformer_blocks.39.audio_attn2.to_v.weight",
1187
+ "transformer_blocks.39.audio_ff.net.0.proj.weight",
1188
+ "transformer_blocks.39.audio_ff.net.2.weight",
1189
+ "transformer_blocks.39.audio_to_video_attn.to_k.weight",
1190
+ "transformer_blocks.39.audio_to_video_attn.to_out.0.weight",
1191
+ "transformer_blocks.39.audio_to_video_attn.to_q.weight",
1192
+ "transformer_blocks.39.audio_to_video_attn.to_v.weight",
1193
+ "transformer_blocks.39.ff.net.0.proj.weight",
1194
+ "transformer_blocks.39.ff.net.2.weight",
1195
+ "transformer_blocks.39.video_to_audio_attn.to_k.weight",
1196
+ "transformer_blocks.39.video_to_audio_attn.to_out.0.weight",
1197
+ "transformer_blocks.39.video_to_audio_attn.to_q.weight",
1198
+ "transformer_blocks.39.video_to_audio_attn.to_v.weight",
1199
+ "transformer_blocks.40.attn1.to_out.0.weight",
1200
+ "transformer_blocks.40.attn1.to_q.weight",
1201
+ "transformer_blocks.40.attn2.to_q.weight",
1202
+ "transformer_blocks.40.audio_attn1.to_k.weight",
1203
+ "transformer_blocks.40.audio_attn1.to_out.0.weight",
1204
+ "transformer_blocks.40.audio_attn1.to_q.weight",
1205
+ "transformer_blocks.40.audio_attn1.to_v.weight",
1206
+ "transformer_blocks.40.audio_attn2.to_k.weight",
1207
+ "transformer_blocks.40.audio_attn2.to_out.0.weight",
1208
+ "transformer_blocks.40.audio_attn2.to_q.weight",
1209
+ "transformer_blocks.40.audio_attn2.to_v.weight",
1210
+ "transformer_blocks.40.audio_ff.net.0.proj.weight",
1211
+ "transformer_blocks.40.audio_ff.net.2.weight",
1212
+ "transformer_blocks.40.audio_to_video_attn.to_k.weight",
1213
+ "transformer_blocks.40.audio_to_video_attn.to_out.0.weight",
1214
+ "transformer_blocks.40.audio_to_video_attn.to_q.weight",
1215
+ "transformer_blocks.40.audio_to_video_attn.to_v.weight",
1216
+ "transformer_blocks.40.ff.net.0.proj.weight",
1217
+ "transformer_blocks.40.ff.net.2.weight",
1218
+ "transformer_blocks.40.video_to_audio_attn.to_out.0.weight",
1219
+ "transformer_blocks.40.video_to_audio_attn.to_q.weight",
1220
+ "transformer_blocks.40.video_to_audio_attn.to_v.weight",
1221
+ "transformer_blocks.41.attn1.to_out.0.weight",
1222
+ "transformer_blocks.41.attn1.to_q.weight",
1223
+ "transformer_blocks.41.attn2.to_q.weight",
1224
+ "transformer_blocks.41.audio_attn1.to_k.weight",
1225
+ "transformer_blocks.41.audio_attn1.to_out.0.weight",
1226
+ "transformer_blocks.41.audio_attn1.to_q.weight",
1227
+ "transformer_blocks.41.audio_attn2.to_k.weight",
1228
+ "transformer_blocks.41.audio_attn2.to_out.0.weight",
1229
+ "transformer_blocks.41.audio_attn2.to_q.weight",
1230
+ "transformer_blocks.41.audio_attn2.to_v.weight",
1231
+ "transformer_blocks.41.audio_ff.net.0.proj.weight",
1232
+ "transformer_blocks.41.audio_ff.net.2.weight",
1233
+ "transformer_blocks.41.audio_to_video_attn.to_k.weight",
1234
+ "transformer_blocks.41.audio_to_video_attn.to_out.0.weight",
1235
+ "transformer_blocks.41.audio_to_video_attn.to_q.weight",
1236
+ "transformer_blocks.41.ff.net.0.proj.weight",
1237
+ "transformer_blocks.41.ff.net.2.weight",
1238
+ "transformer_blocks.41.video_to_audio_attn.to_out.0.weight",
1239
+ "transformer_blocks.41.video_to_audio_attn.to_q.weight",
1240
+ "transformer_blocks.41.video_to_audio_attn.to_v.weight",
1241
+ "transformer_blocks.42.attn1.to_out.0.weight",
1242
+ "transformer_blocks.42.attn1.to_q.weight",
1243
+ "transformer_blocks.42.attn2.to_q.weight",
1244
+ "transformer_blocks.42.audio_attn1.to_k.weight",
1245
+ "transformer_blocks.42.audio_attn1.to_out.0.weight",
1246
+ "transformer_blocks.42.audio_attn1.to_q.weight",
1247
+ "transformer_blocks.42.audio_attn1.to_v.weight",
1248
+ "transformer_blocks.42.audio_attn2.to_k.weight",
1249
+ "transformer_blocks.42.audio_attn2.to_out.0.weight",
1250
+ "transformer_blocks.42.audio_attn2.to_q.weight",
1251
+ "transformer_blocks.42.audio_attn2.to_v.weight",
1252
+ "transformer_blocks.42.audio_ff.net.0.proj.weight",
1253
+ "transformer_blocks.42.audio_to_video_attn.to_k.weight",
1254
+ "transformer_blocks.42.audio_to_video_attn.to_out.0.weight",
1255
+ "transformer_blocks.42.audio_to_video_attn.to_q.weight",
1256
+ "transformer_blocks.42.ff.net.0.proj.weight",
1257
+ "transformer_blocks.42.ff.net.2.weight",
1258
+ "transformer_blocks.42.video_to_audio_attn.to_out.0.weight",
1259
+ "transformer_blocks.42.video_to_audio_attn.to_q.weight",
1260
+ "transformer_blocks.42.video_to_audio_attn.to_v.weight",
1261
+ "transformer_blocks.43.attn1.to_out.0.weight",
1262
+ "transformer_blocks.43.attn1.to_q.weight",
1263
+ "transformer_blocks.43.attn1.to_v.weight",
1264
+ "transformer_blocks.43.attn2.to_q.weight",
1265
+ "transformer_blocks.43.audio_attn1.to_k.weight",
1266
+ "transformer_blocks.43.audio_attn1.to_out.0.weight",
1267
+ "transformer_blocks.43.audio_attn1.to_q.weight",
1268
+ "transformer_blocks.43.audio_attn1.to_v.weight",
1269
+ "transformer_blocks.43.audio_attn2.to_k.weight",
1270
+ "transformer_blocks.43.audio_attn2.to_out.0.weight",
1271
+ "transformer_blocks.43.audio_attn2.to_q.weight",
1272
+ "transformer_blocks.43.audio_attn2.to_v.weight",
1273
+ "transformer_blocks.43.audio_ff.net.0.proj.weight",
1274
+ "transformer_blocks.43.audio_to_video_attn.to_k.weight",
1275
+ "transformer_blocks.43.audio_to_video_attn.to_out.0.weight",
1276
+ "transformer_blocks.43.audio_to_video_attn.to_q.weight",
1277
+ "transformer_blocks.43.audio_to_video_attn.to_v.weight",
1278
+ "transformer_blocks.43.ff.net.0.proj.weight",
1279
+ "transformer_blocks.43.ff.net.2.weight",
1280
+ "transformer_blocks.43.video_to_audio_attn.to_out.0.weight",
1281
+ "transformer_blocks.43.video_to_audio_attn.to_q.weight",
1282
+ "transformer_blocks.43.video_to_audio_attn.to_v.weight",
1283
+ "transformer_blocks.44.attn1.to_out.0.weight",
1284
+ "transformer_blocks.44.attn1.to_q.weight",
1285
+ "transformer_blocks.44.attn1.to_v.weight",
1286
+ "transformer_blocks.44.audio_attn1.to_k.weight",
1287
+ "transformer_blocks.44.audio_attn1.to_q.weight",
1288
+ "transformer_blocks.44.audio_attn1.to_v.weight",
1289
+ "transformer_blocks.44.attn2.to_q.weight",
1290
+ "transformer_blocks.44.audio_attn1.to_out.0.weight",
1291
+ "transformer_blocks.44.audio_attn2.to_k.weight",
1292
+ "transformer_blocks.44.audio_attn2.to_out.0.weight",
1293
+ "transformer_blocks.44.audio_attn2.to_q.weight",
1294
+ "transformer_blocks.44.audio_attn2.to_v.weight",
1295
+ "transformer_blocks.44.audio_ff.net.0.proj.weight",
1296
+ "transformer_blocks.44.audio_to_video_attn.to_k.weight",
1297
+ "transformer_blocks.44.audio_to_video_attn.to_out.0.weight",
1298
+ "transformer_blocks.44.audio_to_video_attn.to_q.weight",
1299
+ "transformer_blocks.44.audio_to_video_attn.to_v.weight",
1300
+ "transformer_blocks.44.ff.net.0.proj.weight",
1301
+ "transformer_blocks.44.ff.net.2.weight",
1302
+ "transformer_blocks.44.video_to_audio_attn.to_out.0.weight",
1303
+ "transformer_blocks.44.video_to_audio_attn.to_q.weight",
1304
+ "transformer_blocks.44.video_to_audio_attn.to_v.weight",
1305
+ "transformer_blocks.45.attn1.to_out.0.weight",
1306
+ "transformer_blocks.45.attn1.to_q.weight",
1307
+ "transformer_blocks.45.attn2.to_q.weight",
1308
+ "transformer_blocks.45.audio_attn1.to_k.weight",
1309
+ "transformer_blocks.45.audio_attn1.to_out.0.weight",
1310
+ "transformer_blocks.45.audio_attn1.to_q.weight",
1311
+ "transformer_blocks.45.audio_attn1.to_v.weight",
1312
+ "transformer_blocks.45.audio_attn2.to_k.weight",
1313
+ "transformer_blocks.45.audio_attn2.to_out.0.weight",
1314
+ "transformer_blocks.45.audio_attn2.to_q.weight",
1315
+ "transformer_blocks.45.audio_attn2.to_v.weight",
1316
+ "transformer_blocks.45.audio_ff.net.0.proj.weight",
1317
+ "transformer_blocks.45.audio_to_video_attn.to_k.weight",
1318
+ "transformer_blocks.45.audio_to_video_attn.to_out.0.weight",
1319
+ "transformer_blocks.45.audio_to_video_attn.to_q.weight",
1320
+ "transformer_blocks.45.audio_to_video_attn.to_v.weight",
1321
+ "transformer_blocks.45.ff.net.0.proj.weight",
1322
+ "transformer_blocks.45.ff.net.2.weight",
1323
+ "transformer_blocks.45.video_to_audio_attn.to_out.0.weight",
1324
+ "transformer_blocks.45.video_to_audio_attn.to_q.weight",
1325
+ "transformer_blocks.45.video_to_audio_attn.to_v.weight",
1326
+ "transformer_blocks.46.attn1.to_out.0.weight",
1327
+ "transformer_blocks.46.attn1.to_q.weight",
1328
+ "transformer_blocks.46.attn2.to_q.weight",
1329
+ "transformer_blocks.46.audio_attn1.to_out.0.weight",
1330
+ "transformer_blocks.46.audio_attn1.to_q.weight",
1331
+ "transformer_blocks.46.audio_attn2.to_k.weight",
1332
+ "transformer_blocks.46.audio_attn2.to_out.0.weight",
1333
+ "transformer_blocks.46.audio_attn2.to_q.weight",
1334
+ "transformer_blocks.46.audio_attn2.to_v.weight",
1335
+ "transformer_blocks.46.audio_ff.net.0.proj.weight",
1336
+ "transformer_blocks.46.audio_to_video_attn.to_out.0.weight",
1337
+ "transformer_blocks.46.audio_to_video_attn.to_q.weight",
1338
+ "transformer_blocks.46.ff.net.0.proj.weight",
1339
+ "transformer_blocks.46.ff.net.2.weight",
1340
+ "transformer_blocks.46.video_to_audio_attn.to_out.0.weight",
1341
+ "transformer_blocks.46.video_to_audio_attn.to_q.weight",
1342
+ "transformer_blocks.46.video_to_audio_attn.to_v.weight",
1343
+ "transformer_blocks.47.attn1.to_out.0.weight",
1344
+ "transformer_blocks.47.attn2.to_k.weight",
1345
+ "transformer_blocks.47.attn2.to_q.weight",
1346
+ "transformer_blocks.47.audio_attn1.to_k.weight",
1347
+ "transformer_blocks.47.audio_attn1.to_out.0.weight",
1348
+ "transformer_blocks.47.audio_attn1.to_q.weight",
1349
+ "transformer_blocks.47.audio_attn2.to_k.weight",
1350
+ "transformer_blocks.47.audio_attn2.to_q.weight",
1351
+ "transformer_blocks.47.audio_attn2.to_v.weight",
1352
+ "transformer_blocks.47.audio_ff.net.0.proj.weight",
1353
+ "transformer_blocks.47.audio_to_video_attn.to_k.weight",
1354
+ "transformer_blocks.47.audio_to_video_attn.to_out.0.weight",
1355
+ "transformer_blocks.47.ff.net.0.proj.weight",
1356
+ "transformer_blocks.47.video_to_audio_attn.to_out.0.weight",
1357
+ "transformer_blocks.47.video_to_audio_attn.to_q.weight"
1358
+ ]
1359
+ },
1360
+ "modules_to_not_convert": [
1361
+ "proj_in",
1362
+ "av_cross_attn_video_scale_shift",
1363
+ "time_embed",
1364
+ "av_cross_attn_video_a2v_gate",
1365
+ "audio_time_embed",
1366
+ "audio_proj_out",
1367
+ "av_cross_attn_audio_v2a_gate",
1368
+ "proj_out",
1369
+ "av_cross_attn_audio_scale_shift",
1370
+ "audio_caption_projection",
1371
+ "audio_proj_in",
1372
+ "caption_projection"
1373
+ ],
1374
+ "non_blocking": false,
1375
+ "quant_conv": false,
1376
+ "quant_method": "sdnq",
1377
+ "quantization_device": null,
1378
+ "quantized_matmul_dtype": "int8",
1379
+ "return_device": null,
1380
+ "sdnq_version": "0.1.4",
1381
+ "svd_rank": 32,
1382
+ "svd_steps": 8,
1383
+ "use_dynamic_quantization": true,
1384
+ "use_grad_ckpt": true,
1385
+ "use_quantized_matmul": false,
1386
+ "use_quantized_matmul_conv": false,
1387
+ "use_static_quantization": true,
1388
+ "use_stochastic_rounding": false,
1389
+ "use_svd": false,
1390
+ "weights_dtype": "int4"
1391
+ }
vae/config.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "AutoencoderKLLTX2Video",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "/home/disty/.cache/huggingface/hub/models--Lightricks--LTX-2/snapshots/64f3f196f34e3deb03c85333ec34e4c9d9f82593/vae",
5
+ "block_out_channels": [
6
+ 256,
7
+ 512,
8
+ 1024,
9
+ 2048
10
+ ],
11
+ "decoder_block_out_channels": [
12
+ 256,
13
+ 512,
14
+ 1024
15
+ ],
16
+ "decoder_causal": false,
17
+ "decoder_inject_noise": [
18
+ false,
19
+ false,
20
+ false,
21
+ false
22
+ ],
23
+ "decoder_layers_per_block": [
24
+ 5,
25
+ 5,
26
+ 5,
27
+ 5
28
+ ],
29
+ "decoder_spatial_padding_mode": "reflect",
30
+ "decoder_spatio_temporal_scaling": [
31
+ true,
32
+ true,
33
+ true
34
+ ],
35
+ "down_block_types": [
36
+ "LTX2VideoDownBlock3D",
37
+ "LTX2VideoDownBlock3D",
38
+ "LTX2VideoDownBlock3D",
39
+ "LTX2VideoDownBlock3D"
40
+ ],
41
+ "downsample_type": [
42
+ "spatial",
43
+ "temporal",
44
+ "spatiotemporal",
45
+ "spatiotemporal"
46
+ ],
47
+ "encoder_causal": true,
48
+ "encoder_spatial_padding_mode": "zeros",
49
+ "in_channels": 3,
50
+ "latent_channels": 128,
51
+ "layers_per_block": [
52
+ 4,
53
+ 6,
54
+ 6,
55
+ 2,
56
+ 2
57
+ ],
58
+ "out_channels": 3,
59
+ "patch_size": 4,
60
+ "patch_size_t": 1,
61
+ "resnet_norm_eps": 1e-06,
62
+ "scaling_factor": 1.0,
63
+ "spatial_compression_ratio": 32,
64
+ "spatio_temporal_scaling": [
65
+ true,
66
+ true,
67
+ true,
68
+ true
69
+ ],
70
+ "temporal_compression_ratio": 8,
71
+ "timestep_conditioning": false,
72
+ "upsample_factor": [
73
+ 2,
74
+ 2,
75
+ 2
76
+ ],
77
+ "upsample_residual": [
78
+ true,
79
+ true,
80
+ true
81
+ ]
82
+ }
vae/diffusion_pytorch_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:107cc359e3c4bce18c53d98686f4b3fe10c4207b6665d89b38b0741270514bfb
3
+ size 2444982370
vocoder/config.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "LTX2Vocoder",
3
+ "_diffusers_version": "0.37.0.dev0",
4
+ "_name_or_path": "/home/disty/.cache/huggingface/hub/models--Lightricks--LTX-2/snapshots/64f3f196f34e3deb03c85333ec34e4c9d9f82593/vocoder",
5
+ "hidden_channels": 1024,
6
+ "in_channels": 128,
7
+ "leaky_relu_negative_slope": 0.1,
8
+ "out_channels": 2,
9
+ "output_sampling_rate": 24000,
10
+ "resnet_dilations": [
11
+ [
12
+ 1,
13
+ 3,
14
+ 5
15
+ ],
16
+ [
17
+ 1,
18
+ 3,
19
+ 5
20
+ ],
21
+ [
22
+ 1,
23
+ 3,
24
+ 5
25
+ ]
26
+ ],
27
+ "resnet_kernel_sizes": [
28
+ 3,
29
+ 7,
30
+ 11
31
+ ],
32
+ "upsample_factors": [
33
+ 6,
34
+ 5,
35
+ 2,
36
+ 2,
37
+ 2
38
+ ],
39
+ "upsample_kernel_sizes": [
40
+ 16,
41
+ 15,
42
+ 8,
43
+ 4,
44
+ 4
45
+ ]
46
+ }
vocoder/diffusion_pytorch_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15855fc59233b9cac50bdd1f0d2ccea4a5eaedbd7fd7549b16d5ebd6cc47d92a
3
+ size 111204124