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
8554e5d
·
verified ·
1 Parent(s): 1078816

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-209000/config.json +79 -0
  3. checkpoint-209000/generation_config.json +6 -0
  4. checkpoint-209000/model.safetensors +3 -0
  5. checkpoint-209000/optimizer.pt +3 -0
  6. checkpoint-209000/rng_state_0.pth +3 -0
  7. checkpoint-209000/rng_state_1.pth +3 -0
  8. checkpoint-209000/rng_state_2.pth +3 -0
  9. checkpoint-209000/rng_state_3.pth +3 -0
  10. checkpoint-209000/rng_state_4.pth +3 -0
  11. checkpoint-209000/rng_state_5.pth +3 -0
  12. checkpoint-209000/rng_state_6.pth +3 -0
  13. checkpoint-209000/rng_state_7.pth +3 -0
  14. checkpoint-209000/scheduler.pt +3 -0
  15. checkpoint-209000/trainer_state.json +3 -0
  16. checkpoint-209000/training_args.bin +3 -0
  17. checkpoint-210000/config.json +79 -0
  18. checkpoint-210000/generation_config.json +6 -0
  19. checkpoint-210000/model.safetensors +3 -0
  20. checkpoint-210000/optimizer.pt +3 -0
  21. checkpoint-210000/rng_state_0.pth +3 -0
  22. checkpoint-210000/rng_state_1.pth +3 -0
  23. checkpoint-210000/rng_state_2.pth +3 -0
  24. checkpoint-210000/rng_state_3.pth +3 -0
  25. checkpoint-210000/rng_state_4.pth +3 -0
  26. checkpoint-210000/rng_state_5.pth +3 -0
  27. checkpoint-210000/rng_state_6.pth +3 -0
  28. checkpoint-210000/rng_state_7.pth +3 -0
  29. checkpoint-210000/scheduler.pt +3 -0
  30. checkpoint-210000/trainer_state.json +3 -0
  31. checkpoint-210000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -70,3 +70,5 @@ checkpoint-193000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
70
  checkpoint-194000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
71
  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
 
 
 
70
  checkpoint-194000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
71
  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
checkpoint-209000/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-209000/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-209000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fac9ed60eb294b39142ba93ee430cb28b733b0b6e1856c18186c764710e5956d
3
+ size 4411759432
checkpoint-209000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5329426c9acef912c3edefb197d25e9b2ce2387b38cab03e204b07fb99db7e0
3
+ size 6332050591
checkpoint-209000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:989474f1d9df33077e5673e03a37c851ba20df33cc51204a3800623b76677be4
3
+ size 16389
checkpoint-209000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f6115a0d7375a7d6c069e2ad98d4d685c1355ba6aade85171538019ff47354f
3
+ size 16389
checkpoint-209000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79377f6ef250eea20982ea621df5fab5129a769cfcce3030feaa85fb3a772abe
3
+ size 16389
checkpoint-209000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8df88ea94dca038118b46c47dd888ba3421f86a498d64959bae5eb21d3f10d5
3
+ size 16389
checkpoint-209000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80a825d624ff7aa85a6945769909b5115e3a1c49cc38094c237117a0de17e693
3
+ size 16389
checkpoint-209000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6659a9c438ec2b2b5d2bf43b12b1355025df23b09584d56cf482af49a0622bb
3
+ size 16389
checkpoint-209000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee4650a99d9189b1acfb91aa5ae19a8c9f16937064d01ea2b92b9a8be5e81ae1
3
+ size 16389
checkpoint-209000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12718beecf7e0fa670aa6d1ae5153e1b1757d27aa8f140376852aa1acec828bd
3
+ size 16389
checkpoint-209000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8333c5fc38d14210b197d74edb8b05f37061c4aec0d0b7d484b88cb256cb876c
3
+ size 1465
checkpoint-209000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:272df356386e93ceb2959b67fac6dda93ea7f43dd70516c634791e8a9b870247
3
+ size 32154152
checkpoint-209000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebeee5639d8a83d0eb627319b1001e776dd5a48cf83df39fb9c1617a3c1a1baf
3
+ size 6481
checkpoint-210000/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-210000/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-210000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55ea3755654a8e9728048083b710e8f48777d3cfaa7e110d77029cc8cb12e489
3
+ size 4411759432
checkpoint-210000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:087cc9f7d6dfd8a5322c9d30662ee16e3c41431798e966242f3e14137790f53f
3
+ size 6332050591
checkpoint-210000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff4bbea5185a98076d590877d208bb1ca1e140fcf5f147965b3fad15b8569636
3
+ size 16389
checkpoint-210000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c960b0971957487a3b5d56fd46e5dfcc68719017b2960e31bc6285ae1a114ae
3
+ size 16389
checkpoint-210000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ec835a0be1454b0319debfd17e46728ee22b67b586b1e82ec28ca432c054a86
3
+ size 16389
checkpoint-210000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3941ebc729d49658e704252988c0d0abc609d0dafe0fd0445a4656af6a66e82f
3
+ size 16389
checkpoint-210000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c8d45f1937aa3451c77c58550f66ad3f3147e2c9e1fe86be46a556d672e283a
3
+ size 16389
checkpoint-210000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f7e58e10328e6fc29da8c4760e8691848dc3082698d639f96c197faa7177c43
3
+ size 16389
checkpoint-210000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94f8e9b78a725d810f1d56d4214e2288c3e141a95820ab73b83eb925b8f4fd56
3
+ size 16389
checkpoint-210000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba239816b25aef5b87380cd59113f2df1511cc76e4714d73ff31db68f9ec0745
3
+ size 16389
checkpoint-210000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c0fc6f465aa55de71cce6abc20c60ffd14e2fc3fb9dcf30f446744d341f769e
3
+ size 1465
checkpoint-210000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7889abd5a13c85fb07400d7e12161c6c1e061b2fb1e7b3d23951e3fd03fc7951
3
+ size 32308515
checkpoint-210000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ebeee5639d8a83d0eb627319b1001e776dd5a48cf83df39fb9c1617a3c1a1baf
3
+ size 6481