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
2a95247
·
verified ·
1 Parent(s): 33a997a

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-131000/config.json +79 -0
  3. checkpoint-131000/generation_config.json +6 -0
  4. checkpoint-131000/model.safetensors +3 -0
  5. checkpoint-131000/optimizer.pt +3 -0
  6. checkpoint-131000/rng_state_0.pth +3 -0
  7. checkpoint-131000/rng_state_1.pth +3 -0
  8. checkpoint-131000/rng_state_2.pth +3 -0
  9. checkpoint-131000/rng_state_3.pth +3 -0
  10. checkpoint-131000/rng_state_4.pth +3 -0
  11. checkpoint-131000/rng_state_5.pth +3 -0
  12. checkpoint-131000/rng_state_6.pth +3 -0
  13. checkpoint-131000/rng_state_7.pth +3 -0
  14. checkpoint-131000/scheduler.pt +3 -0
  15. checkpoint-131000/trainer_state.json +3 -0
  16. checkpoint-131000/training_args.bin +3 -0
  17. checkpoint-132000/config.json +79 -0
  18. checkpoint-132000/generation_config.json +6 -0
  19. checkpoint-132000/model.safetensors +3 -0
  20. checkpoint-132000/optimizer.pt +3 -0
  21. checkpoint-132000/rng_state_0.pth +3 -0
  22. checkpoint-132000/rng_state_1.pth +3 -0
  23. checkpoint-132000/rng_state_2.pth +3 -0
  24. checkpoint-132000/rng_state_3.pth +3 -0
  25. checkpoint-132000/rng_state_4.pth +3 -0
  26. checkpoint-132000/rng_state_5.pth +3 -0
  27. checkpoint-132000/rng_state_6.pth +3 -0
  28. checkpoint-132000/rng_state_7.pth +3 -0
  29. checkpoint-132000/scheduler.pt +3 -0
  30. checkpoint-132000/trainer_state.json +3 -0
  31. checkpoint-132000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -52,3 +52,5 @@ checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
52
  checkpoint-128000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
53
  checkpoint-129000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
52
  checkpoint-128000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
53
  checkpoint-129000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
+ checkpoint-131000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
+ checkpoint-132000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-131000/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-131000/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-131000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:272970ca72d02cf05128567e15cf424e0f87bb30147eb51a20871930040aa1ea
3
+ size 4411759432
checkpoint-131000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c7b25ef8cb2276619b79f584505a761f210282804a2fcaf71acef2b42567aa4
3
+ size 6332050591
checkpoint-131000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84cf9a840fc24d68783b840c0ea955e1d802e11972fccf87157e235422498c12
3
+ size 16389
checkpoint-131000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9c03eec53e8258ebf4b4d31f98b94f0e8601bf99096a4c616e85d931c40ca3e
3
+ size 16389
checkpoint-131000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fab175228bcda009a32c57d871e48768bdfc1c82dc7adb4f76ecae2810e261f
3
+ size 16389
checkpoint-131000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8d92e057b762f76893b8c41de19b6cc24aa4a774cae31e7037b72db6cf7b82e
3
+ size 16389
checkpoint-131000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62e30a0c500e500df480247d46482c07b8988f4895d80e31c987861f1306e722
3
+ size 16389
checkpoint-131000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62b8df5f46ffb4e0197dbc504af63cf3e170c1f0e3c8f1159c9927707b87a2d0
3
+ size 16389
checkpoint-131000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3e9ea4c084a5f26f10f8956985aaec98506fde7f65b49255d50ee94d343c09c
3
+ size 16389
checkpoint-131000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b74632035ef41bb4bf8f97678d2da1d53e5d9f312b68a7bd0abe01bede3ea2e
3
+ size 16389
checkpoint-131000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:717902f819098ee4bd0a6eb70a6b2c7dd5bc2f52ce9a2c6325a65a5c6d0dd734
3
+ size 1465
checkpoint-131000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e0ae8bc94566561c940d3936b80493fa799f825ec2f347c2a8040488a17435b
3
+ size 20106010
checkpoint-131000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e7a7243197dd963f39bce225f40e21f97143793760b132f227c444418f9e7e4
3
+ size 6481
checkpoint-132000/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-132000/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-132000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c19d3488f87200598f7e793ed12ecced6252d490a108807c9ab7668b0eda7fd
3
+ size 4411759432
checkpoint-132000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3c09abe462d93f50b8384baff400c867642f9c9fe213bdee2253b0a339b58e8
3
+ size 6332050591
checkpoint-132000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48d181b14c34e270b5aff8c839c4f4e13a1ecfdee3a731aabb74d38ce75879d4
3
+ size 16389
checkpoint-132000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b3a564f4180dac38ce3a6de0600d9e124b2bc4b3235fb2e1d54b1c6fac339d4
3
+ size 16389
checkpoint-132000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c584ead737f1d5c752b0633a11b5459e906beb8c2018bbd5ca6dcc338a3f8565
3
+ size 16389
checkpoint-132000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:528d6f0aa8d1f6b7129327e22a17adbfe1009064be3d712b799f2f658f9e2e44
3
+ size 16389
checkpoint-132000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59394be60693fdeaf3241194318cd8b3a233651beed3ca39bedd61ee36d7e6e4
3
+ size 16389
checkpoint-132000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:353ffacc1fc3459c8c24cfc363783d06086d20aadc7f128c2c488490862ac64f
3
+ size 16389
checkpoint-132000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28f0f25a8846c263b5531e1a3c8d9145f042fd8af3781d90b23a444f9b5afe16
3
+ size 16389
checkpoint-132000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc1203adef13dacdfdfda52ec0e2d7776f584405950da0e6773007d72fa60de6
3
+ size 16389
checkpoint-132000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c20509369fa849c7f9a0f4dc7c2c0148b87a8eb59856974050bdd32d8b2cf244
3
+ size 1465
checkpoint-132000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27aaf8f5a404d1ea4313e70a989399926b91c7e3c5e313f1da199c726d19ceab
3
+ size 20260663
checkpoint-132000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7b9897afd7104fe65e590877a80c1f7e73fcb61e07388fb015fe42397f1ddb3
3
+ size 6417