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
79ebdfb
·
verified ·
1 Parent(s): 7552f67

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-11000/config.json +79 -0
  2. checkpoint-11000/generation_config.json +6 -0
  3. checkpoint-11000/model.safetensors +3 -0
  4. checkpoint-11000/optimizer.pt +3 -0
  5. checkpoint-11000/rng_state_0.pth +3 -0
  6. checkpoint-11000/rng_state_1.pth +3 -0
  7. checkpoint-11000/rng_state_2.pth +3 -0
  8. checkpoint-11000/rng_state_3.pth +3 -0
  9. checkpoint-11000/rng_state_4.pth +3 -0
  10. checkpoint-11000/rng_state_5.pth +3 -0
  11. checkpoint-11000/rng_state_6.pth +3 -0
  12. checkpoint-11000/rng_state_7.pth +3 -0
  13. checkpoint-11000/scheduler.pt +3 -0
  14. checkpoint-11000/trainer_state.json +0 -0
  15. checkpoint-11000/training_args.bin +3 -0
  16. checkpoint-12000/config.json +79 -0
  17. checkpoint-12000/generation_config.json +6 -0
  18. checkpoint-12000/model.safetensors +3 -0
  19. checkpoint-12000/optimizer.pt +3 -0
  20. checkpoint-12000/rng_state_0.pth +3 -0
  21. checkpoint-12000/rng_state_1.pth +3 -0
  22. checkpoint-12000/rng_state_2.pth +3 -0
  23. checkpoint-12000/rng_state_3.pth +3 -0
  24. checkpoint-12000/rng_state_4.pth +3 -0
  25. checkpoint-12000/rng_state_5.pth +3 -0
  26. checkpoint-12000/rng_state_6.pth +3 -0
  27. checkpoint-12000/rng_state_7.pth +3 -0
  28. checkpoint-12000/scheduler.pt +3 -0
  29. checkpoint-12000/trainer_state.json +0 -0
  30. checkpoint-12000/training_args.bin +3 -0
checkpoint-11000/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-11000/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-11000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59efb478c3129bb9d5df8a3b7436883bc05043d468e1c296bffaf54930de9be3
3
+ size 4411759432
checkpoint-11000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:754b6f8082a0d5e61a524939e48ea2db99f278c028dd87554ed430e8909a71b5
3
+ size 6332050591
checkpoint-11000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d600c9ffed94bcc567f96e9500e486b485e5b381dcf420580a0b68087e96034
3
+ size 16389
checkpoint-11000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d20707679f26f5bd5e9a39e02e660605324d6f95cd9a1896bcfaac106ae017c
3
+ size 16389
checkpoint-11000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bdb232c0b59697149e052b9107ef64c6318f69b1243b57c869d6e615481da7a
3
+ size 16389
checkpoint-11000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:266d0e6d687de9e9d35da8266b105c9e875249b1c51dda63553af8564cc8a1df
3
+ size 16389
checkpoint-11000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0213ae6843fea7d897ffb10be27363b213a07a69dd0eaac2bbaec3c4deb64e49
3
+ size 16389
checkpoint-11000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc35fc055aa3ef346e322b39cb2e3dc9a2cf4f93bf92e8e8d891e95f93d9788e
3
+ size 16389
checkpoint-11000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2589862b91fd9ea07e8ea0a2bd230aabe15bb524994c539f9ad29a26a809ade2
3
+ size 16389
checkpoint-11000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0af37f1a8ce30c2ebeb9696374767ff50fbdc140bf26ebc6054e2b08647f6989
3
+ size 16389
checkpoint-11000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0575d6271ea94ed81fb39c8b390d8d4a70785154518f81220088c9b9726a3425
3
+ size 1465
checkpoint-11000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-11000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac1babf2e9dd0d17c127a6ec11f72284f82280a2bef442ebfa51dd129f9e2e13
3
+ size 6417
checkpoint-12000/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-12000/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-12000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42c6c7f516dd7b796082a5ccdb5e05031f51a074624b694270f31a75c89d005b
3
+ size 4411759432
checkpoint-12000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abadba4da6b5e0da6b3e1c198132aba05e40e96f6c9c074961931a67cd2da74c
3
+ size 6332050591
checkpoint-12000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c10a85abc476677700cc1ccf2c89605f7bffd6f33df54c08e34b63409245fb2
3
+ size 16389
checkpoint-12000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f705c601a890fd1139c4b19c0fbf7183c4468b38c71bca7a9e00b1005d9d1e0d
3
+ size 16389
checkpoint-12000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:571bdedab555f8bbea8fd2955fa332feab4ceb501fbb91651bdf17c1dbb24f61
3
+ size 16389
checkpoint-12000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db28a1d9cb0e2f122fda0263010113c7a36894d20961452c4c47b6cad17d7919
3
+ size 16389
checkpoint-12000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a35cd3fc94d3c48cc600144edcefcf3e6d18a252ff48ec144ca38d12a311a9c4
3
+ size 16389
checkpoint-12000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:870b0d2a588b7888533c08486e5da402926ea65e1b091b40e3e2b8999ed0dda9
3
+ size 16389
checkpoint-12000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:686aeb28e6a8d55b0bb4b30c5f691f1e12cd22e7e6bbf4f5ad57aa397e495292
3
+ size 16389
checkpoint-12000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7e9e2126e4385b04824f07de74fdbe0de2bed95735c912f05e8531f4a4e9967
3
+ size 16389
checkpoint-12000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1958f565f1327527f6a12edff8e7e4c04f1bdae2678439e4c0a0f7696b262464
3
+ size 1465
checkpoint-12000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-12000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ac418064fd5ffb1673ceb40eca7bff26f67237a046147a3ad362c55adceb205
3
+ size 6417