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
ed4ae35
·
verified ·
1 Parent(s): 230b5b9

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-100000/config.json +79 -0
  3. checkpoint-100000/generation_config.json +6 -0
  4. checkpoint-100000/model.safetensors +3 -0
  5. checkpoint-100000/optimizer.pt +3 -0
  6. checkpoint-100000/rng_state_0.pth +3 -0
  7. checkpoint-100000/rng_state_1.pth +3 -0
  8. checkpoint-100000/rng_state_2.pth +3 -0
  9. checkpoint-100000/rng_state_3.pth +3 -0
  10. checkpoint-100000/rng_state_4.pth +3 -0
  11. checkpoint-100000/rng_state_5.pth +3 -0
  12. checkpoint-100000/rng_state_6.pth +3 -0
  13. checkpoint-100000/rng_state_7.pth +3 -0
  14. checkpoint-100000/scheduler.pt +3 -0
  15. checkpoint-100000/trainer_state.json +3 -0
  16. checkpoint-100000/training_args.bin +3 -0
  17. checkpoint-99000/config.json +79 -0
  18. checkpoint-99000/generation_config.json +6 -0
  19. checkpoint-99000/model.safetensors +3 -0
  20. checkpoint-99000/optimizer.pt +3 -0
  21. checkpoint-99000/rng_state_0.pth +3 -0
  22. checkpoint-99000/rng_state_1.pth +3 -0
  23. checkpoint-99000/rng_state_2.pth +3 -0
  24. checkpoint-99000/rng_state_3.pth +3 -0
  25. checkpoint-99000/rng_state_4.pth +3 -0
  26. checkpoint-99000/rng_state_5.pth +3 -0
  27. checkpoint-99000/rng_state_6.pth +3 -0
  28. checkpoint-99000/rng_state_7.pth +3 -0
  29. checkpoint-99000/scheduler.pt +3 -0
  30. checkpoint-99000/trainer_state.json +3 -0
  31. checkpoint-99000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -43,3 +43,5 @@ checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-89000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  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
 
 
 
43
  checkpoint-89000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  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
checkpoint-100000/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-100000/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-100000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f150eba407db2f88dc18b67fdbe6776fdc813acc38dbf432f31dca355a2e3e6c
3
+ size 4411759432
checkpoint-100000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46d421f94b8b80188c608e5030ca6c8e14a33c141fe372aa80a20b25cfde2ee0
3
+ size 6332050591
checkpoint-100000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a0177cde181417e7f00ed8efe71c3c0c143ae4cd54b67c30feff6df9eed6cc8
3
+ size 16389
checkpoint-100000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50cf44ca7d39e9cfa22aac090fdfa961dfd944d94b11d68bdfb9fb2928ef2b20
3
+ size 16389
checkpoint-100000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c904454123492e9ab92d36ceacffa60cc50fed3e00e4da2fa8621372471b7b19
3
+ size 16389
checkpoint-100000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c97ea2fc53b299c732ef191dc6ad3c8ad7ebeb9960211b98cb1d3443e6a080b8
3
+ size 16389
checkpoint-100000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd555fca28aa651801a0ba094a20de880db56c1883bf8a900e8fa594b1576b16
3
+ size 16389
checkpoint-100000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fd30ecf75f9b58e0d89c31c579282137a2c731835271153e76b42a799da0c71
3
+ size 16389
checkpoint-100000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b0a4ef08342aea646b3dd6c982f574027faca2104856a4942b440a0026d855e
3
+ size 16389
checkpoint-100000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c72b1a2043a565f157d196bd6a5ec87b1591e7b46b89ba5735f6f03a7d43298
3
+ size 16389
checkpoint-100000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ac99a084bcaaf43dbe0be3b3f6a0442c66bb00c70b940a0f70ec0fedc2481d3
3
+ size 1465
checkpoint-100000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7be3cab7c79ced75f019f8eeee168085a890772fab58e4429284fa4503546eab
3
+ size 15312019
checkpoint-100000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b6650b3126decb450270769b236f936eb481d45d350bf0c0bb3ae49feb750da
3
+ size 6481
checkpoint-99000/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-99000/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-99000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fe8351e3a716254d16590229d3b9340a37b7c327eca186f92951987dae9e196
3
+ size 4411759432
checkpoint-99000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:556b620ccbf995895a79f8f281310c9f3b0e8cfdeb1b3e76090989adcaa2f462
3
+ size 6332050591
checkpoint-99000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51831182850799e48334ed53776dd509f42a8bbf80bf72ac34c516b7e262868d
3
+ size 16389
checkpoint-99000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d606d62fb5be46329e372556a3693a2f1b4a7ae2e00dabc74850697ee1102426
3
+ size 16389
checkpoint-99000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f86d62e20d8eb032fd1d57815a9522b9e478ec2ee76447fbe679974630f20f19
3
+ size 16389
checkpoint-99000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34e4e9039244e6f1164fb0da0645f20a0287980d0dc205341e0b6dc638da8267
3
+ size 16389
checkpoint-99000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:306b04203ebf3a0039ccaac790ac307f130a86b1a35b5a3daf50841a23d10d69
3
+ size 16389
checkpoint-99000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb9dd3415206ffcc4e62393629bcb875e92e005064cefd2a4edcdae5fea90ecb
3
+ size 16389
checkpoint-99000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cd2dbbf10beb59a04444ebff5413362f4c7178127793903714298e5ca1c7168
3
+ size 16389
checkpoint-99000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57974d9e4d77a2ba0ebcee07fea6fa8fd10aebfb6fae4be0030878da138d6e3d
3
+ size 16389
checkpoint-99000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64727631f04465abaeca5fe0eb1a02d37024ae3ed46a27ebf58e311e17808202
3
+ size 1465
checkpoint-99000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d18447ce37d92338f6b4962213f5dff22e59de36fe318df2730b387c1518c15e
3
+ size 15158312
checkpoint-99000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b6650b3126decb450270769b236f936eb481d45d350bf0c0bb3ae49feb750da
3
+ size 6481