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
12f1af8
·
verified ·
1 Parent(s): eb00cdc

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-119000/config.json +79 -0
  3. checkpoint-119000/generation_config.json +6 -0
  4. checkpoint-119000/model.safetensors +3 -0
  5. checkpoint-119000/optimizer.pt +3 -0
  6. checkpoint-119000/rng_state_0.pth +3 -0
  7. checkpoint-119000/rng_state_1.pth +3 -0
  8. checkpoint-119000/rng_state_2.pth +3 -0
  9. checkpoint-119000/rng_state_3.pth +3 -0
  10. checkpoint-119000/rng_state_4.pth +3 -0
  11. checkpoint-119000/rng_state_5.pth +3 -0
  12. checkpoint-119000/rng_state_6.pth +3 -0
  13. checkpoint-119000/rng_state_7.pth +3 -0
  14. checkpoint-119000/scheduler.pt +3 -0
  15. checkpoint-119000/trainer_state.json +3 -0
  16. checkpoint-119000/training_args.bin +3 -0
  17. checkpoint-120000/config.json +79 -0
  18. checkpoint-120000/generation_config.json +6 -0
  19. checkpoint-120000/model.safetensors +3 -0
  20. checkpoint-120000/optimizer.pt +3 -0
  21. checkpoint-120000/rng_state_0.pth +3 -0
  22. checkpoint-120000/rng_state_1.pth +3 -0
  23. checkpoint-120000/rng_state_2.pth +3 -0
  24. checkpoint-120000/rng_state_3.pth +3 -0
  25. checkpoint-120000/rng_state_4.pth +3 -0
  26. checkpoint-120000/rng_state_5.pth +3 -0
  27. checkpoint-120000/rng_state_6.pth +3 -0
  28. checkpoint-120000/rng_state_7.pth +3 -0
  29. checkpoint-120000/scheduler.pt +3 -0
  30. checkpoint-120000/trainer_state.json +3 -0
  31. checkpoint-120000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -47,3 +47,5 @@ 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
 
 
 
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
50
+ checkpoint-119000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
51
+ checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-119000/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-119000/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-119000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d960126e66bd47a7a8bd3f0af395a25ab7689ed0e57ffbff61416c6e74945122
3
+ size 4411759432
checkpoint-119000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cec6e4e1e5b8548dfa21b5489af7ff4a3f1b24759f2627575aff74e0bac9ce0e
3
+ size 6332050591
checkpoint-119000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bd314d7d63b72d93098f8de274074ff624993be489761d1604661082b8f7210
3
+ size 16389
checkpoint-119000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc975ceef782bfff0a95690428c1e340c25706e3b44cd7e0b281a79bf61bc5b4
3
+ size 16389
checkpoint-119000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90c8153c6c6b201134239024fd40f9b8160d02ead35a054b61edad8bfa937088
3
+ size 16389
checkpoint-119000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7065ed781284a13bb5d7ca32cfa8fd09e1699e0bc4b480084da24a982e7dcdaf
3
+ size 16389
checkpoint-119000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f515f94104e0da46efa46d47455e0fd8f50c8117b237f2f76e73f8fc69ca8afa
3
+ size 16389
checkpoint-119000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20ab964c5741418795e45cc2b702fa4439075495e975b616a23a31652e550092
3
+ size 16389
checkpoint-119000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e78780807c52ec986d23c49562c3b09eaaa2e7373a555c2823cc3db684ad9d04
3
+ size 16389
checkpoint-119000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:262c5054a6f52faaee28ec4d5fe9d793ae6ef369c9000ca13dd1fe548186332f
3
+ size 16389
checkpoint-119000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9da0985b066318a9701641b07e6d41e14b81be81b1ddd43838e36b59f75e57d
3
+ size 1465
checkpoint-119000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1ca08b8531ba7db41c4fb3fad8d0728064d3a42932474d332bdb2da1cfcaf15
3
+ size 18250208
checkpoint-119000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:543bfbab331abf4e74aabe87ad5c7e8f3ed3eef8c6b6d2777de7178440c85098
3
+ size 6481
checkpoint-120000/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-120000/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-120000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f555a9b307a00bea5bbe278e4a54a40a739b7835081d49d61fa414288943260
3
+ size 4411759432
checkpoint-120000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33795115899e1bdbcbbbd08357c09f49d436e6a23638635a815d71fe854e6f36
3
+ size 6332050591
checkpoint-120000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:285734927e9aedf5f66b95fd93d718e6e8f50a9bca603f8fb32c59d977f3089b
3
+ size 16389
checkpoint-120000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d4c60ebfd59a492e471456c782876c02e44b0f666a26e185fd1aeb88c50c3f4
3
+ size 16389
checkpoint-120000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68eb36ea7036efe5f9d009d4d40bec15f86fedf609a41cbe2e4da9375e2b1048
3
+ size 16389
checkpoint-120000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dc95d7ea9219d3107946359e73b2b1bdbfc71aaa7567663ebb4844a42f5b5d9
3
+ size 16389
checkpoint-120000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9368d8056d3811921710746c556653558b4f1658223c9e059b3fed3bd6e0cf58
3
+ size 16389
checkpoint-120000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5f0d2b7f101a04a5ba081e59d247ec9392de744fab2f87f2789a84f14b058d0
3
+ size 16389
checkpoint-120000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0f99cbd5d7e51fb3c26b5ad5382ba0cc42ce9d0bd982a45dc1729338a8223fe
3
+ size 16389
checkpoint-120000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cb45393af1e66e046d260247d91fd58f1c8fb1fa94276a80eef019e0721d931
3
+ size 16389
checkpoint-120000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da42563ae985ca2ada0fb55cd0c7d2997223340fbd020b8bf146cb107422a679
3
+ size 1465
checkpoint-120000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:168b582d54210551949dbf44db93164e5299bd669476776c71449d76226d6912
3
+ size 18404905
checkpoint-120000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:543bfbab331abf4e74aabe87ad5c7e8f3ed3eef8c6b6d2777de7178440c85098
3
+ size 6481