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
e4eaa96
·
verified ·
1 Parent(s): 313958b

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-179000/config.json +79 -0
  3. checkpoint-179000/generation_config.json +6 -0
  4. checkpoint-179000/model.safetensors +3 -0
  5. checkpoint-179000/optimizer.pt +3 -0
  6. checkpoint-179000/rng_state_0.pth +3 -0
  7. checkpoint-179000/rng_state_1.pth +3 -0
  8. checkpoint-179000/rng_state_2.pth +3 -0
  9. checkpoint-179000/rng_state_3.pth +3 -0
  10. checkpoint-179000/rng_state_4.pth +3 -0
  11. checkpoint-179000/rng_state_5.pth +3 -0
  12. checkpoint-179000/rng_state_6.pth +3 -0
  13. checkpoint-179000/rng_state_7.pth +3 -0
  14. checkpoint-179000/scheduler.pt +3 -0
  15. checkpoint-179000/trainer_state.json +3 -0
  16. checkpoint-179000/training_args.bin +3 -0
  17. checkpoint-180000/config.json +79 -0
  18. checkpoint-180000/generation_config.json +6 -0
  19. checkpoint-180000/model.safetensors +3 -0
  20. checkpoint-180000/optimizer.pt +3 -0
  21. checkpoint-180000/rng_state_0.pth +3 -0
  22. checkpoint-180000/rng_state_1.pth +3 -0
  23. checkpoint-180000/rng_state_2.pth +3 -0
  24. checkpoint-180000/rng_state_3.pth +3 -0
  25. checkpoint-180000/rng_state_4.pth +3 -0
  26. checkpoint-180000/rng_state_5.pth +3 -0
  27. checkpoint-180000/rng_state_6.pth +3 -0
  28. checkpoint-180000/rng_state_7.pth +3 -0
  29. checkpoint-180000/scheduler.pt +3 -0
  30. checkpoint-180000/trainer_state.json +3 -0
  31. checkpoint-180000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -62,3 +62,5 @@ checkpoint-159000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
62
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
+ checkpoint-179000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
66
+ checkpoint-180000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-179000/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-179000/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-179000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:917351ce186023dc9d6f35249a3de729748ba48cee3edb4be0852ec9d3e6c082
3
+ size 4411759432
checkpoint-179000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c596ee3a1adb843426bcd3bd065eec260500e6a57285f5a5b9bd54eb45588860
3
+ size 6332050591
checkpoint-179000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4441530228d8f8a0ffb91fc8635c622635fd4976ebed0f651e21d7c793d574f
3
+ size 16389
checkpoint-179000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8799f6fd487822332c62907703bc2d035d7efc56f87f1e27047caae426f47f65
3
+ size 16389
checkpoint-179000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2da131c03ca8748cd5a000eef4d885b099733dc759daa9194c5169a52c7e72c9
3
+ size 16389
checkpoint-179000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aac3fbad28c82b6bead260a7828b4dec744fbbb7e04c9a67b66128fe3792d49
3
+ size 16389
checkpoint-179000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f511e5176ac51727ca3036c47e494c1b6c7770d0de3ed36c405e6677bc5309e2
3
+ size 16389
checkpoint-179000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d7a54662a4b1cf55b8c1132e6b39138b396d0de6d014a720f6fee3561bbb673
3
+ size 16389
checkpoint-179000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e385c8b3011a258e7e73588ce3526a23b031ca0fa979ff0852d50081d974bb27
3
+ size 16389
checkpoint-179000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89973be48efc2b624df5da8b91ca54858cb848cd2331d9100424855150d5e7a2
3
+ size 16389
checkpoint-179000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e9c3020b6395ab91d90bfc20ddfa4e6e4eeeaae4b4bfcadbb4ad5e1a0959ab9f
3
+ size 1465
checkpoint-179000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14de5d74ac826e4cf06ed3277c1dc5ab3c91201b920bc97c0d66a0066a89331d
3
+ size 27523510
checkpoint-179000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59d0ea813943498cda3fa511fd79f976233930037c88dfd11170a0698c080690
3
+ size 6481
checkpoint-180000/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-180000/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-180000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a55302aaa70b3ee16258bf283e3d1c677f16fb137b2fba84badfbd0e1035e757
3
+ size 4411759432
checkpoint-180000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93a3b92a3d702fd06f0d0171c73d7e6be9aaec93da5f5e38f114fb5230a694f2
3
+ size 6332050591
checkpoint-180000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9d405fcbdcdbf1a3720be098334000d2ac25ff5aafdd2bbb4a0338c2ba9fb44
3
+ size 16389
checkpoint-180000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e8938b7325529e369bf75036c4aef2986e0652e11125077a5f714225a0d7267
3
+ size 16389
checkpoint-180000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6dff1ac73eaf432bb4e83a9af7e4b5e73ff918aa54a0bf61057f060d9d9d9b73
3
+ size 16389
checkpoint-180000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5764a81c512ddb9fb638b7910d223fe6301e9eb45dcf3ed85ca4eaf4360f2d86
3
+ size 16389
checkpoint-180000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd315a29e6b9ac031af35a5ac5cf47990b75b479af1c70264bc0f1a8f578d77d
3
+ size 16389
checkpoint-180000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90926bee1a72eb794671c0c648aca454f6f55b89c9697f451ed6c34321a077b9
3
+ size 16389
checkpoint-180000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e88ef6b40b73eee808e4b3a53df483e39bf9ebf9531a91109a5382983f1cefff
3
+ size 16389
checkpoint-180000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a14c95803f79a82b51496616e737ee84900470ed3d5ce0248f4627d4558bf12
3
+ size 16389
checkpoint-180000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20dbbb08dd09cce94cd6015c6eaedccba081602b9c24823d44ecaf238508e3fc
3
+ size 1465
checkpoint-180000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa370fcd72f8f111b4e8368fa8fb29a23fdeb9480b95e0d6b0d50b9e51c6aee6
3
+ size 27677927
checkpoint-180000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59d0ea813943498cda3fa511fd79f976233930037c88dfd11170a0698c080690
3
+ size 6481