Hugging Face's logo

umd-vt-nyu
/
JH_multinode_dc-ae-f32c32-in-1.0-diffusers-768_patch-1_group-7_5e4_residual_attn

Safetensors
Model card Files
xet
Community
jiuhai commited on
Commit
938c670
·
verified ·
1 Parent(s): 9397731

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-49000/config.json +79 -0
  2. checkpoint-49000/generation_config.json +6 -0
  3. checkpoint-49000/model.safetensors +3 -0
  4. checkpoint-49000/optimizer.pt +3 -0
  5. checkpoint-49000/rng_state_0.pth +3 -0
  6. checkpoint-49000/rng_state_1.pth +3 -0
  7. checkpoint-49000/rng_state_2.pth +3 -0
  8. checkpoint-49000/rng_state_3.pth +3 -0
  9. checkpoint-49000/rng_state_4.pth +3 -0
  10. checkpoint-49000/rng_state_5.pth +3 -0
  11. checkpoint-49000/rng_state_6.pth +3 -0
  12. checkpoint-49000/rng_state_7.pth +3 -0
  13. checkpoint-49000/scheduler.pt +3 -0
  14. checkpoint-49000/trainer_state.json +0 -0
  15. checkpoint-49000/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-49000/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": 1,
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-49000/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-49000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7be5a8b27bf5661e814068362b50510402c7dbd164293dc74be4ed566d7acc92
3
+ size 4411759432
checkpoint-49000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78425321f74f4fe9e65ad42e89142c224636cc12af0654acdd05e0cccab9ddcd
3
+ size 6332050591
checkpoint-49000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff02a01f3477924689c38a9d75c66a63258b27e0226e64b9e0b6a7f8dfb34bbe
3
+ size 16389
checkpoint-49000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2dd170261e2703757b1b59d07ab35c60374312466652bdd7061f09a29c7afabc
3
+ size 16389
checkpoint-49000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c01d33bb17f61fcdc07b2e12635d29fe17c39274c78111f9db51c8288fd19a4
3
+ size 16389
checkpoint-49000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b39e3efc6be11d44dc7a320ada4046d2769c226acc5342fc47dffed2d8a54cfd
3
+ size 16389
checkpoint-49000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8126bd4593bff359af59a100721c08f366689f3a10a6bdd6ebda42e3d3d1e84
3
+ size 16389
checkpoint-49000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75ba9a54617380331fb7df982069a1b40a450809d75a5324381d83317e7ea12b
3
+ size 16389
checkpoint-49000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a117d22f264843b23d3578ee24e4fc69dab3874fa0afddb0b100d1942f6208dd
3
+ size 16389
checkpoint-49000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c375d5a4796d52ea4a7c3f8ea786e35ceeec5cabeb600c96b0c842d02fd87e09
3
+ size 16389
checkpoint-49000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77614b8385f34ae8e441dcffe5113db415f8ceacab6b76a1d0fb449b528c848d
3
+ size 1465
checkpoint-49000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-49000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:728d72c03905fffec5d34b566354f58cf88c03a8eec89617065c7facedbf1643
3
+ size 6481
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": 1,
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 sha256:27a7f54655853a6a39b23c8dd6fd4a6bdb912d484cb0cd8c9bfc1ad1a3a7d2b6
3
+ size 4411759432
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5166ccbc971763328ef4c23aed9542d609e9ed329da91a5e1ecefef79f7da41
3
+ size 6332050591
checkpoint-50000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d7d0e0bce2051a747c685998a75f3de1b6a4a08e3695fa6dacb67b066043865
3
+ size 16389
checkpoint-50000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9838c4964e54b3ccad025c4007c17f203e59b18e7ea2e2507824e5de1c36c4a2
3
+ size 16389
checkpoint-50000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9feea813989bb0192f25c3e2f6b456ce6a998fc7a3c9daa0287e069bfeb53831
3
+ size 16389
checkpoint-50000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec32edff8a51b914e9e5c0226f8a4b199a017b7b0a2b7ab0be8d35cf553953f2
3
+ size 16389
checkpoint-50000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e4ad5d66b29766879d524c268e9eecf411d9a9304e933418e0c3af32708ad6a
3
+ size 16389
checkpoint-50000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f682523b01b0818b0f36273c7b60f7e500ecedabdd39b461c58217021b529d03
3
+ size 16389
checkpoint-50000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa5daa9b386e816f2ca35680afc1394d706a8cb9b19616a0b4b8ab3e98e99e0c
3
+ size 16389
checkpoint-50000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fd76342ba3fe4f588c412093eefc9519d3501eac8dccae3734792388d69eaf3
3
+ size 16389
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4019c9b4563050e244cba02d8431666fa9b7fc59a0224040bdb75c6c936fa4b5
3
+ size 1465
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 sha256:728d72c03905fffec5d34b566354f58cf88c03a8eec89617065c7facedbf1643
3
+ size 6481