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 Community
jiuhai commited on
Commit
fd8f14e
·
verified ·
1 Parent(s): a486ebf

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-259000/config.json +79 -0
  3. checkpoint-259000/generation_config.json +6 -0
  4. checkpoint-259000/model.safetensors +3 -0
  5. checkpoint-259000/optimizer.pt +3 -0
  6. checkpoint-259000/rng_state_0.pth +3 -0
  7. checkpoint-259000/rng_state_1.pth +3 -0
  8. checkpoint-259000/rng_state_2.pth +3 -0
  9. checkpoint-259000/rng_state_3.pth +3 -0
  10. checkpoint-259000/rng_state_4.pth +3 -0
  11. checkpoint-259000/rng_state_5.pth +3 -0
  12. checkpoint-259000/rng_state_6.pth +3 -0
  13. checkpoint-259000/rng_state_7.pth +3 -0
  14. checkpoint-259000/scheduler.pt +3 -0
  15. checkpoint-259000/trainer_state.json +3 -0
  16. checkpoint-259000/training_args.bin +3 -0
  17. checkpoint-260000/config.json +79 -0
  18. checkpoint-260000/generation_config.json +6 -0
  19. checkpoint-260000/model.safetensors +3 -0
  20. checkpoint-260000/optimizer.pt +3 -0
  21. checkpoint-260000/rng_state_0.pth +3 -0
  22. checkpoint-260000/rng_state_1.pth +3 -0
  23. checkpoint-260000/rng_state_2.pth +3 -0
  24. checkpoint-260000/rng_state_3.pth +3 -0
  25. checkpoint-260000/rng_state_4.pth +3 -0
  26. checkpoint-260000/rng_state_5.pth +3 -0
  27. checkpoint-260000/rng_state_6.pth +3 -0
  28. checkpoint-260000/rng_state_7.pth +3 -0
  29. checkpoint-260000/scheduler.pt +3 -0
  30. checkpoint-260000/trainer_state.json +3 -0
  31. checkpoint-260000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -80,3 +80,5 @@ checkpoint-239000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-249000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
80
  checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-249000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
83
+ checkpoint-259000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
84
+ checkpoint-260000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-259000/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-259000/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-259000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:853e1eb4a535d18059b8dd95fc2efae8b9371634292aa504716165056a9fae24
3
+ size 4411759432
checkpoint-259000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a47d926cf74b0b50b8e29ee1fd19c82f0be6922d32990ac91cb08171b4a77ab6
3
+ size 6332050591
checkpoint-259000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d9295f84847bf2416f6eaa5c922c7933603b40dd2aa8f63a8391da2662b528e
3
+ size 16389
checkpoint-259000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83f66dc968e6514884aa24b2c0501f72dcb10b8f4b7d7777c2eea87b7d72dfc1
3
+ size 16389
checkpoint-259000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:800c6511b70ef54d91d79c36e48a163dddeb6d30b7a3960c4dc1a6a6f82e95b5
3
+ size 16389
checkpoint-259000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:836342d0c0244ea02a9460a0141352f1bb7953d39a2ef42a1669f661ddd31747
3
+ size 16389
checkpoint-259000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a34e00ae94c2cd94dccfdd6eebd85dc710a11eab21e3a2c256642bc8d8909fb
3
+ size 16389
checkpoint-259000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5224a887f5312fc9d151f567d6e274b075c212c2e9a6a10dd64b1177c46be0b9
3
+ size 16389
checkpoint-259000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbafaa6481cd4c39142428465405b3dd02a27517323e1ef779c63604593f439f
3
+ size 16389
checkpoint-259000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ac6e3a03defe876ee5be34f0ce35b5d662978894b3095303e8f75f9e5be09fb
3
+ size 16389
checkpoint-259000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36ad0b4816b3cacfd0564ac82186a040c9a3efcb8362b2c66e3baeb326713ad8
3
+ size 1465
checkpoint-259000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b9ca2d8ba0527f741c44776204c8e2173e517592d3bc684fb192d55dc444a8d
3
+ size 39871456
checkpoint-259000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:115ef29b0e59c8e5b1e4dcef0551952e744b0bc9a45ac094e0ad1b5970a820ca
3
+ size 6481
checkpoint-260000/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-260000/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-260000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff5132e36748e532505ddf973dbedd4d8a58a187e5856ce22a46f0001b30187c
3
+ size 4411759432
checkpoint-260000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab0a29debc495812ffa59d8823caedfc5916262e3be69ba10617eee8e481d837
3
+ size 6332050591
checkpoint-260000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b23b34bb1d9457838c78d42fc9c09eb4d327f3ddfb15ac1e43b7afeb56db4b3
3
+ size 16389
checkpoint-260000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6f78390a0e7edb9b7ed912a2ca854a06178a350f48cc277e5e0dca48ed5a18d
3
+ size 16389
checkpoint-260000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd7a5684c48dd55f35dec90aa496375a5d3b5dc91558f75ad68f21c38fd8bfbd
3
+ size 16389
checkpoint-260000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d4231e4379d6dcd5d0933a85fcff39ad274037c18cd819906f1ded5e05e3806
3
+ size 16389
checkpoint-260000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ab8c482cd8ba9a34691511ca93f49b35fc2f417a0d0a820123b57f1742ad658
3
+ size 16389
checkpoint-260000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7b19952f46a4152c7f0b5f7bbf7c99b7d76de056c9b3c9370aee99c6d1f3fcc
3
+ size 16389
checkpoint-260000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f915e01c132994cbbeffc189393d4bc5ef2ccdda4fccc1c910916a93b4a0947
3
+ size 16389
checkpoint-260000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fde301fe203442506cfad894ad15992a337da3caef5911983b94b0d24eda065
3
+ size 16389
checkpoint-260000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6d72cfa72982a9ce1baefbfb4d1a7521d6f8914d5ebf0f927290dea5e4110c8
3
+ size 1465
checkpoint-260000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d903fa633de8ff241df029ad8726527967860ae289f78584047efeedb701311
3
+ size 40025766
checkpoint-260000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:115ef29b0e59c8e5b1e4dcef0551952e744b0bc9a45ac094e0ad1b5970a820ca
3
+ size 6481