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
97d4998
·
verified ·
1 Parent(s): 8554e5d

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-219000/config.json +79 -0
  3. checkpoint-219000/generation_config.json +6 -0
  4. checkpoint-219000/model.safetensors +3 -0
  5. checkpoint-219000/optimizer.pt +3 -0
  6. checkpoint-219000/rng_state_0.pth +3 -0
  7. checkpoint-219000/rng_state_1.pth +3 -0
  8. checkpoint-219000/rng_state_2.pth +3 -0
  9. checkpoint-219000/rng_state_3.pth +3 -0
  10. checkpoint-219000/rng_state_4.pth +3 -0
  11. checkpoint-219000/rng_state_5.pth +3 -0
  12. checkpoint-219000/rng_state_6.pth +3 -0
  13. checkpoint-219000/rng_state_7.pth +3 -0
  14. checkpoint-219000/scheduler.pt +3 -0
  15. checkpoint-219000/trainer_state.json +3 -0
  16. checkpoint-219000/training_args.bin +3 -0
  17. checkpoint-220000/config.json +79 -0
  18. checkpoint-220000/generation_config.json +6 -0
  19. checkpoint-220000/model.safetensors +3 -0
  20. checkpoint-220000/optimizer.pt +3 -0
  21. checkpoint-220000/rng_state_0.pth +3 -0
  22. checkpoint-220000/rng_state_1.pth +3 -0
  23. checkpoint-220000/rng_state_2.pth +3 -0
  24. checkpoint-220000/rng_state_3.pth +3 -0
  25. checkpoint-220000/rng_state_4.pth +3 -0
  26. checkpoint-220000/rng_state_5.pth +3 -0
  27. checkpoint-220000/rng_state_6.pth +3 -0
  28. checkpoint-220000/rng_state_7.pth +3 -0
  29. checkpoint-220000/scheduler.pt +3 -0
  30. checkpoint-220000/trainer_state.json +3 -0
  31. checkpoint-220000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -72,3 +72,5 @@ checkpoint-199000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
72
  checkpoint-200000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
73
  checkpoint-209000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
74
  checkpoint-210000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
72
  checkpoint-200000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
73
  checkpoint-209000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
74
  checkpoint-210000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
75
+ checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
+ checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-219000/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-219000/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-219000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7af49bd1bdbd01494d4a866906e462c47e5d23579a04d52a91cac169af3653a4
3
+ size 4411759432
checkpoint-219000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f708984f6812a86bb41c420a9f547307de6cfcc9bdbabfb3a36a98dda960377
3
+ size 6332050591
checkpoint-219000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3a47562d52c26f13625bd9b965c2af614a4c3c47f6dace8afaeb18c8adcec8d
3
+ size 16389
checkpoint-219000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cbbb637c778816877498f4a9d36e21d28bc59a9d9a25bcddbd8185bc611c95c
3
+ size 16389
checkpoint-219000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c570c6e1ff5596de64fa59cab0d53bfdc48bad59b609dde92fe3d4c60bb1d6ea
3
+ size 16389
checkpoint-219000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60fa61428c66aec033b2b30fe3804894509ca008174d7246ea6e2ef6a5c78fc3
3
+ size 16389
checkpoint-219000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:456ee04f2e57881425b38b1439efddf0dd4e312dd857ca00c030950a50943708
3
+ size 16389
checkpoint-219000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eeb2c8ba35007690e90ce1bc611936f068875c5ed750d9a498c62f67567ad5e
3
+ size 16389
checkpoint-219000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61e93ceae21784790e0b66759f68c7849b4f02e8a7b56caa0973bbb84b7157fc
3
+ size 16389
checkpoint-219000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff67280c0f3db6024b1d0e6ad77fb9ba832e2f39cb13f5caceb6a29d7e069f43
3
+ size 16389
checkpoint-219000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc1e9b2f46b4901deec6fa912ff3f8233f0bc2c02f43111300a91980737b6464
3
+ size 1465
checkpoint-219000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fd2518b288a51b4c2e820bfeafbcec1b716546fd557774c3714eab4b2a52885
3
+ size 33697692
checkpoint-219000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f228cb129441c6eac3e3a4facbbb5be866aead336cc92c95b8966ff8f3201dcb
3
+ size 6481
checkpoint-220000/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-220000/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-220000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06795767b0b12b0f4c1eae91497133f645eba70a0138e004cf305ec6d97013e3
3
+ size 4411759432
checkpoint-220000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:556085dd9c0ed88d4388380cee5992b6c9068a80516a4e4a2abcd8f62015e979
3
+ size 6332050591
checkpoint-220000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e67860f4fd4768237d6210a9eb3216e172d5d313f58b213146b21d69754444f1
3
+ size 16389
checkpoint-220000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0e2148a6d32a63f005c22beddf11366507b5b260df06a6e5d0af4bbd0162f9f
3
+ size 16389
checkpoint-220000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ec971f39ac6ddfd0b0be78c68b37857e2cf23ccbd24d3bf87c9878a8edf4116
3
+ size 16389
checkpoint-220000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8fa85b2d5c7461d7f38ea46c07878bc7b0526f5fc8aa607fbddddf476c7c6f1
3
+ size 16389
checkpoint-220000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ec0f1423ef9a52967121cde14635d28dc451892821e9b15aea579dad4970536
3
+ size 16389
checkpoint-220000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c132acbb4f97faf9925e788f65456e3e715741923d493d4aa991180592a776ca
3
+ size 16389
checkpoint-220000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:211607347a5e926159bac820db85cd343cf63a8c7d76b6ae26c1887e3eb8c0d1
3
+ size 16389
checkpoint-220000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e76110dfd5ad3dcfc1c41c3f1d65df2008e64639d99987765afb44c09ee0c3c1
3
+ size 16389
checkpoint-220000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2f83f2cca52030ce7ab3f9028254498fd674f00e9e640dbf485afccefc366f2
3
+ size 1465
checkpoint-220000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6b8c069cd53f090523ec5b4089d038a63e3b8e18c0cdaa1b810abe39c0e1dd0
3
+ size 33852041
checkpoint-220000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f228cb129441c6eac3e3a4facbbb5be866aead336cc92c95b8966ff8f3201dcb
3
+ size 6481