Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
b93b05b
·
verified ·
1 Parent(s): 0ff83ab

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-38000/config.json +79 -0
  2. checkpoint-38000/generation_config.json +6 -0
  3. checkpoint-38000/model.safetensors +3 -0
  4. checkpoint-38000/optimizer.pt +3 -0
  5. checkpoint-38000/rng_state_0.pth +3 -0
  6. checkpoint-38000/rng_state_1.pth +3 -0
  7. checkpoint-38000/rng_state_10.pth +3 -0
  8. checkpoint-38000/rng_state_11.pth +3 -0
  9. checkpoint-38000/rng_state_12.pth +3 -0
  10. checkpoint-38000/rng_state_13.pth +3 -0
  11. checkpoint-38000/rng_state_14.pth +3 -0
  12. checkpoint-38000/rng_state_15.pth +3 -0
  13. checkpoint-38000/rng_state_2.pth +3 -0
  14. checkpoint-38000/rng_state_3.pth +3 -0
  15. checkpoint-38000/rng_state_4.pth +3 -0
  16. checkpoint-38000/rng_state_5.pth +3 -0
  17. checkpoint-38000/rng_state_6.pth +3 -0
  18. checkpoint-38000/rng_state_7.pth +3 -0
  19. checkpoint-38000/rng_state_8.pth +3 -0
  20. checkpoint-38000/rng_state_9.pth +3 -0
  21. checkpoint-38000/scheduler.pt +3 -0
  22. checkpoint-38000/trainer_state.json +0 -0
  23. checkpoint-38000/training_args.bin +3 -0
  24. checkpoint-39000/config.json +79 -0
  25. checkpoint-39000/generation_config.json +6 -0
  26. checkpoint-39000/model.safetensors +3 -0
  27. checkpoint-39000/optimizer.pt +3 -0
  28. checkpoint-39000/rng_state_0.pth +3 -0
  29. checkpoint-39000/rng_state_1.pth +3 -0
  30. checkpoint-39000/rng_state_10.pth +3 -0
  31. checkpoint-39000/rng_state_11.pth +3 -0
  32. checkpoint-39000/rng_state_12.pth +3 -0
  33. checkpoint-39000/rng_state_13.pth +3 -0
  34. checkpoint-39000/rng_state_14.pth +3 -0
  35. checkpoint-39000/rng_state_15.pth +3 -0
  36. checkpoint-39000/rng_state_2.pth +3 -0
  37. checkpoint-39000/rng_state_3.pth +3 -0
  38. checkpoint-39000/rng_state_4.pth +3 -0
  39. checkpoint-39000/rng_state_5.pth +3 -0
  40. checkpoint-39000/rng_state_6.pth +3 -0
  41. checkpoint-39000/rng_state_7.pth +3 -0
  42. checkpoint-39000/rng_state_8.pth +3 -0
  43. checkpoint-39000/rng_state_9.pth +3 -0
  44. checkpoint-39000/scheduler.pt +3 -0
  45. checkpoint-39000/trainer_state.json +0 -0
  46. checkpoint-39000/training_args.bin +3 -0
checkpoint-38000/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": 2,
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-sana-1.1-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-38000/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-38000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90807f212f8801c81c582679dde2def49c7a6ea88b0a4955d9c5ca7d674ae251
3
+ size 4539487822
checkpoint-38000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6902c2dfc1d0b7d99b5dfc2ebb7e7f2e7b7f6d396c64216fbe7d1039e90932aa
3
+ size 6632345482
checkpoint-38000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:851aa924c6ce547a4bb7091422d899a81c4a9f6cb0eec25d4a602d8a73afabae
3
+ size 15984
checkpoint-38000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68cff0d0b89ce87a85194e9608f26d91d97875583065f0d80a8e0c02b71d8ae7
3
+ size 15984
checkpoint-38000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9ee7015df05af2030275833b72571620831dedb2d995c8663ce2ece2d690211
3
+ size 15997
checkpoint-38000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f75cbcf0be88b1754446af3fa12c68d5c4983cc756a9144f32fa624f123b27b
3
+ size 15997
checkpoint-38000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5ae79d7323e10ea338eda8e0a63d9f146bcb812d88263556770c7d75a5a052f
3
+ size 15997
checkpoint-38000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:192cbd2f3e142da7c83ec1df6b1ea6f0bfb74c71213c33145f3a429ca67c727b
3
+ size 15997
checkpoint-38000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc0836eb2d37cdb2641865455cefdc5e96ba937f017a208fe66045152598cf3c
3
+ size 15997
checkpoint-38000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7cc09c672d1d0a9f93787401ae011d8d99b8823425d8b2c3ee8a404598f8e90d
3
+ size 15997
checkpoint-38000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5a3936523d0b4d5b2077fb2e2bc3e6cada1778a043fe15b438c798577ad7829
3
+ size 15984
checkpoint-38000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec80238ca6a405817957afe0486aa1e8533f066f1f04ceca8248d8891cf08bde
3
+ size 15984
checkpoint-38000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2dd22d69042163b7b5b8ff46b5ce5952556b1ed38004c6b7ade390dce1a8b909
3
+ size 15984
checkpoint-38000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1982623be2f24e850db1fb2d9f290b07086cf72df4b95ba577bb46e621017778
3
+ size 15984
checkpoint-38000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74ee8bae774299d55044e06dead819697dd16b64185f7bbe0d65c6819a5df880
3
+ size 15984
checkpoint-38000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:259204b27d122917a9dc934892088f1f190db228e343d7f29009ac65f72628be
3
+ size 15984
checkpoint-38000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b6c34467e56a3ec934bbd9bb03d43f43662bccd8171f5bc69f8a6ba569ef9b9
3
+ size 15984
checkpoint-38000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f44a8c1d58290dd05d104304305d1f700ba915282e65c3a54f0ab7e3367f0f5
3
+ size 15984
checkpoint-38000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7b987c90811de117ba5bf2f292e72dca849fcd63c0aaf7fff0a35abb1ab8e4c
3
+ size 1064
checkpoint-38000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-38000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:adc83beb8878ef0e1d2bc5781640ac2d22762538c79bab213feb9e8246010f0a
3
+ size 6008
checkpoint-39000/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": 2,
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-sana-1.1-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-39000/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-39000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63234ec341c59be32b79c92aba63387823d09f59e910810f8575545e50fa52f4
3
+ size 4539487822
checkpoint-39000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b03507f018b944e62164e5033cf6e3ee0db63512109f07cfdec91b52e9854065
3
+ size 6632345482
checkpoint-39000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:497fa1302b26729ba5b426d0cabb1fbcb85652d1405943263692ff362f8b66ea
3
+ size 15984
checkpoint-39000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:828ba5ee32f87c41c8a23c3020daa5b3c3d2bcf8a5d912ca8838b96fe74b8c06
3
+ size 15984
checkpoint-39000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd7ba4e83be1c50e8ade65e978243029dcaa5337c734a44a0928a595991614fc
3
+ size 15997
checkpoint-39000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97af13ca9aceb53e7a12cd274e0da702dd6b433a2dbe05846e4b0538ee8587c7
3
+ size 15997
checkpoint-39000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bcb3dc20da12c0645d283a48611e940a8a6a128d0e869de94cd12e6c4d2f8013
3
+ size 15997
checkpoint-39000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:717226b77382710d83ecf4b18d34ee5c4bdc98ed1080db333105742d57cb664d
3
+ size 15997
checkpoint-39000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0df39cbc576c0428a3b42269b0455e9f63c15437142c3e5016d16a0e7300e769
3
+ size 15997
checkpoint-39000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdb82e87f86f3de8167f26ed0ff502939007296b84e4eb0cd389289ea6b3f0f0
3
+ size 15997
checkpoint-39000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f373b0041a59ee5d6feef7a1d3fc5718663cc9d48975c2d6956282e6e203bab7
3
+ size 15984
checkpoint-39000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ad605017ae3c4487b3dc9e6477c3a88ec884ab6362862d43797c1f71b793de2
3
+ size 15984
checkpoint-39000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aae5302ad28e604174e027780bb7f0ae4b030d92682cd31d4466edf724562ca5
3
+ size 15984
checkpoint-39000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4df59b217dc7a36aab9b038bff86041ae3af98c434d01acd619a156ff6a0203a
3
+ size 15984
checkpoint-39000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f97f312bc5fc8972235d85377e29160626f8eb5d85dd837d4caa974c3b3bd259
3
+ size 15984
checkpoint-39000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6617ab045be2082121b5a6f517a22cc26d389415f5660c911ed5d1ea3e2202b
3
+ size 15984
checkpoint-39000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a84a93c8751c5888fe4f0e3620959d4be4cefdc3d10eb0b830166f4597ade713
3
+ size 15984
checkpoint-39000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae2208111f041e54093782e03d910c77af7603c691ee183e930700af9d4b2063
3
+ size 15984
checkpoint-39000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d801487764f4e4dfb2f53d6832675284fe07d40d3aace2fb24d20cdf6e5fc3cc
3
+ size 1064
checkpoint-39000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84c88446eb97454f598b779b130e17d776129c752d5ac4aae1f508e9f8efe57e
3
+ size 6008