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
eb00cdc
·
verified ·
1 Parent(s): ed4ae35

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-109000/config.json +79 -0
  3. checkpoint-109000/generation_config.json +6 -0
  4. checkpoint-109000/model.safetensors +3 -0
  5. checkpoint-109000/optimizer.pt +3 -0
  6. checkpoint-109000/rng_state_0.pth +3 -0
  7. checkpoint-109000/rng_state_1.pth +3 -0
  8. checkpoint-109000/rng_state_2.pth +3 -0
  9. checkpoint-109000/rng_state_3.pth +3 -0
  10. checkpoint-109000/rng_state_4.pth +3 -0
  11. checkpoint-109000/rng_state_5.pth +3 -0
  12. checkpoint-109000/rng_state_6.pth +3 -0
  13. checkpoint-109000/rng_state_7.pth +3 -0
  14. checkpoint-109000/scheduler.pt +3 -0
  15. checkpoint-109000/trainer_state.json +3 -0
  16. checkpoint-109000/training_args.bin +3 -0
  17. checkpoint-110000/config.json +79 -0
  18. checkpoint-110000/generation_config.json +6 -0
  19. checkpoint-110000/model.safetensors +3 -0
  20. checkpoint-110000/optimizer.pt +3 -0
  21. checkpoint-110000/rng_state_0.pth +3 -0
  22. checkpoint-110000/rng_state_1.pth +3 -0
  23. checkpoint-110000/rng_state_2.pth +3 -0
  24. checkpoint-110000/rng_state_3.pth +3 -0
  25. checkpoint-110000/rng_state_4.pth +3 -0
  26. checkpoint-110000/rng_state_5.pth +3 -0
  27. checkpoint-110000/rng_state_6.pth +3 -0
  28. checkpoint-110000/rng_state_7.pth +3 -0
  29. checkpoint-110000/scheduler.pt +3 -0
  30. checkpoint-110000/trainer_state.json +3 -0
  31. checkpoint-110000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -45,3 +45,5 @@ checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
45
  checkpoint-91000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
46
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
47
  checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
45
  checkpoint-91000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
46
  checkpoint-100000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
47
  checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
48
+ checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
49
+ checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-109000/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-109000/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-109000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c609d4025ed2329327376089a8995f8265a5623c963ddd639f9d4655c4ab175d
3
+ size 4411759432
checkpoint-109000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b3630462b043a51d7c7bcf2345d6992da2583a9d9be9c0875bcbda8cd8b97e4
3
+ size 6332050591
checkpoint-109000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de0f43025969d917f7d17966eaa8beb9f6832f1a4de325a9eafe6dbe9f1ee71f
3
+ size 16389
checkpoint-109000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ee80759f394fdaf8cce8379255b7b0da2fa7fb2e27738e7909726df2c40cc08
3
+ size 16389
checkpoint-109000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a229a042b8ae0fdbc7df5b4c60638b5b8bbf88bb34780a86109b2f225bceb77
3
+ size 16389
checkpoint-109000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b7e27abfaf2db5ba9b32b0a9c133da2f882c0e54894a9695e84a9a1a202661f
3
+ size 16389
checkpoint-109000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7e64b8d453dcaaac32c3a41ba12229f5694eb43465c053eb8089c4d453ccc43
3
+ size 16389
checkpoint-109000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2513999892ef8108585de81a056e30125ab8370398b08667f1867771b352458d
3
+ size 16389
checkpoint-109000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89660eb174f0bd144a1b0af93c4d05d3da64d87b41ef593567ae78a3eaa37a09
3
+ size 16389
checkpoint-109000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06fe68361b62e07793f5a18080de959641307d3c22114ad6ab000d5f93bae910
3
+ size 16389
checkpoint-109000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18810b0d9881e189a32db77df5c36193bf2797414d230b323aef11b9fc5693b2
3
+ size 1465
checkpoint-109000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e387b5679a67f2214c6ce85f46cad162a9ce68ca9fdcb214c8e43a7eeb93a338
3
+ size 16703831
checkpoint-109000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7aaa3435c9acbe6f501cff69c92a688fe88a9067392078dcaf447bd477a7a9aa
3
+ size 6481
checkpoint-110000/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-110000/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-110000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d2009159282914d60b87d54a1cc3f2173333a284bb5de42f1b33be90ef5abfb
3
+ size 4411759432
checkpoint-110000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f03c7b400b0ce4eb76a4edaca8ba788111af0d830092d08efdd034e56cbfd69
3
+ size 6332050591
checkpoint-110000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e94a05ceb8c730637b6cac53d3473bceb522227b67c631eec0e80877c59d537e
3
+ size 16389
checkpoint-110000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d16f025f4e23c91c935b7af3670ee28efe27c90d8480360342f319bb84b50084
3
+ size 16389
checkpoint-110000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1202e6ff715d0b7d94c1e396005f65f294976ee493080912dab38651d28217a
3
+ size 16389
checkpoint-110000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:749df4fd251f78bef3d3a5156b8216c5890de9463d44d021f412691707c5f2a7
3
+ size 16389
checkpoint-110000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:072aec9c76906c50447715212a237d3f35299bf3301feceed9cb2feddcca6094
3
+ size 16389
checkpoint-110000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63f69cc0810bc6572f3943596452acc9630137728256989a4972f1c60b3e1c79
3
+ size 16389
checkpoint-110000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9fcb22776cdcb3336863cda0ce5867be8f780927df3330e16d8eb13e627b88e
3
+ size 16389
checkpoint-110000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d513ae9a4fb292c41a58c7e268b0a572ced15b2fb80ff2556dcbd55968cd8cb
3
+ size 16389
checkpoint-110000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94ea140bf2a801de89c070bb6f20cfa3418d2f9ac2baa7ad8a2899d6262aae75
3
+ size 1465
checkpoint-110000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d3427ee3886568216ae784feaa31584c98ab2ca2779c416a40111a972627c08
3
+ size 16858482
checkpoint-110000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7aaa3435c9acbe6f501cff69c92a688fe88a9067392078dcaf447bd477a7a9aa
3
+ size 6481