Hugging Face's logo

zhiyang1
/
vlm_dc-ae-f32c32-in-1.0-diffusers-768_patch-2_epoch-64-0_group-7_5e4_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
7f9f2e2
·
verified ·
1 Parent(s): 5f34aaa

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-49500/config.json +79 -0
  2. checkpoint-49500/generation_config.json +6 -0
  3. checkpoint-49500/model.safetensors +3 -0
  4. checkpoint-49500/optimizer.pt +3 -0
  5. checkpoint-49500/rng_state_0.pth +3 -0
  6. checkpoint-49500/rng_state_1.pth +3 -0
  7. checkpoint-49500/rng_state_2.pth +3 -0
  8. checkpoint-49500/rng_state_3.pth +3 -0
  9. checkpoint-49500/rng_state_4.pth +3 -0
  10. checkpoint-49500/rng_state_5.pth +3 -0
  11. checkpoint-49500/rng_state_6.pth +3 -0
  12. checkpoint-49500/rng_state_7.pth +3 -0
  13. checkpoint-49500/scheduler.pt +3 -0
  14. checkpoint-49500/trainer_state.json +0 -0
  15. checkpoint-49500/training_args.bin +3 -0
  16. checkpoint-50000/config.json +79 -0
  17. checkpoint-50000/generation_config.json +6 -0
  18. checkpoint-50000/model.safetensors +3 -0
  19. checkpoint-50000/optimizer.pt +3 -0
  20. checkpoint-50000/rng_state_0.pth +3 -0
  21. checkpoint-50000/rng_state_1.pth +3 -0
  22. checkpoint-50000/rng_state_2.pth +3 -0
  23. checkpoint-50000/rng_state_3.pth +3 -0
  24. checkpoint-50000/rng_state_4.pth +3 -0
  25. checkpoint-50000/rng_state_5.pth +3 -0
  26. checkpoint-50000/rng_state_6.pth +3 -0
  27. checkpoint-50000/rng_state_7.pth +3 -0
  28. checkpoint-50000/scheduler.pt +3 -0
  29. checkpoint-50000/trainer_state.json +0 -0
  30. checkpoint-50000/training_args.bin +3 -0
checkpoint-49500/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 2,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-in-1.0-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-49500/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-49500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b1e603241e8d80114cbbc1e3dd8a99a1a3394271c9f08f5bde9cbff2afadb62
3
+ size 4561722280
checkpoint-49500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba148c96635b3e006706c6df08a9934d5cfa9ddc54c534779db991de6f07115b
3
+ size 6632330442
checkpoint-49500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab8418cc1115c6c62b4a5224a90ba68169c06eafd568249cda57e6f6e8bf62f3
3
+ size 15984
checkpoint-49500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86905968f2a3f9fce715821189e94a7654310be914dd4071c4d9cd400b1f2075
3
+ size 15984
checkpoint-49500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b70ae7f6e2e7a2752f7d0ec38d5701302435e71585421bb8f14a8d108a5d519f
3
+ size 15984
checkpoint-49500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7cac335891e9fdbc3ac07e66cd863fa372b888fd1bb8e0ae3e1d8030f42c142b
3
+ size 15984
checkpoint-49500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8078bdefce96c0e2924f6ee3e81996c86e4f8d01f6512682d03d7d1468faafb4
3
+ size 15984
checkpoint-49500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4dea22a00b05b86f2a77095780d72aa91bb01939408da926232aed2e57a8364e
3
+ size 15984
checkpoint-49500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba7200e577f0e83d62403f50329020ec14b5015857cf615bbc3f272523d1e6a7
3
+ size 15984
checkpoint-49500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a695eae2c8c0d3f4f9664ee3e6d5963a126d2b019431ba393a9f661434d8ee00
3
+ size 15984
checkpoint-49500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec9e2bcc8642a9f8bea0f3001f648ce1fd466072fc28c4819796ed5cacbdcbf9
3
+ size 1064
checkpoint-49500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-49500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4825d5416dd1ad6b66637cac441cea19d6113fda608d23e1ba2e2ba53a77957c
3
+ size 6072
checkpoint-50000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 2,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-in-1.0-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-50000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-50000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57ce38602a0ecf4a6748cc7f2bec7199ad697058a3c21e279658c4d77ae9841d
3
+ size 4561722280
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0372f11066f049fa5bc18ce8dc1fec9016e8e43a7231c21d8c0a5d7bd02dfe96
3
+ size 6632330442
checkpoint-50000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc0d44f8f7c1e100d81c50d5c6696b68ba6e8bd94c186f9aab3aff922f061ec1
3
+ size 15984
checkpoint-50000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f3a34e24686480faa7463a669e1a276926ee337aa54b35289f5629a25cb6978
3
+ size 15984
checkpoint-50000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e399c7a4ca806d44634e21c220c6c0fa271efb4871f70b15c9bb1fd899c150c
3
+ size 15984
checkpoint-50000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a987023be285c580f69c8601999435681ce5020c2f8a6a252b6e2efd0b7d4328
3
+ size 15984
checkpoint-50000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f5b1b398274412f5f7d0d755c8dfebdae998cedd2fa2fbd56ba6f4ef2352646
3
+ size 15984
checkpoint-50000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99c13e1c3d009f185a001463c13f2661c3056c5518cc795c1b7474dd238a8ef8
3
+ size 15984
checkpoint-50000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc6907bd9c4f64e493072b119f7df6bb4338353f3beca9280facb53202d8bb42
3
+ size 15984
checkpoint-50000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1612c29a6aec97fe3fdbcc46563f2fda62cbf1901322f141d320b007dbeac516
3
+ size 15984
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc44d1103e7d1616dd2d2f393a781b7690c295712b013d5889080915890b784f
3
+ size 1064
checkpoint-50000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4825d5416dd1ad6b66637cac441cea19d6113fda608d23e1ba2e2ba53a77957c
3
+ size 6072