Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
f555f71
·
verified ·
1 Parent(s): 6d3b5a4

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-10000/config.json +79 -0
  2. checkpoint-10000/generation_config.json +6 -0
  3. checkpoint-10000/model.safetensors +3 -0
  4. checkpoint-10000/optimizer.pt +3 -0
  5. checkpoint-10000/rng_state_0.pth +3 -0
  6. checkpoint-10000/rng_state_1.pth +3 -0
  7. checkpoint-10000/rng_state_10.pth +3 -0
  8. checkpoint-10000/rng_state_11.pth +3 -0
  9. checkpoint-10000/rng_state_12.pth +3 -0
  10. checkpoint-10000/rng_state_13.pth +3 -0
  11. checkpoint-10000/rng_state_14.pth +3 -0
  12. checkpoint-10000/rng_state_15.pth +3 -0
  13. checkpoint-10000/rng_state_2.pth +3 -0
  14. checkpoint-10000/rng_state_3.pth +3 -0
  15. checkpoint-10000/rng_state_4.pth +3 -0
  16. checkpoint-10000/rng_state_5.pth +3 -0
  17. checkpoint-10000/rng_state_6.pth +3 -0
  18. checkpoint-10000/rng_state_7.pth +3 -0
  19. checkpoint-10000/rng_state_8.pth +3 -0
  20. checkpoint-10000/rng_state_9.pth +3 -0
  21. checkpoint-10000/scheduler.pt +3 -0
  22. checkpoint-10000/trainer_state.json +0 -0
  23. checkpoint-10000/training_args.bin +3 -0
  24. checkpoint-9000/config.json +79 -0
  25. checkpoint-9000/generation_config.json +6 -0
  26. checkpoint-9000/model.safetensors +3 -0
  27. checkpoint-9000/optimizer.pt +3 -0
  28. checkpoint-9000/rng_state_0.pth +3 -0
  29. checkpoint-9000/rng_state_1.pth +3 -0
  30. checkpoint-9000/rng_state_10.pth +3 -0
  31. checkpoint-9000/rng_state_11.pth +3 -0
  32. checkpoint-9000/rng_state_12.pth +3 -0
  33. checkpoint-9000/rng_state_13.pth +3 -0
  34. checkpoint-9000/rng_state_14.pth +3 -0
  35. checkpoint-9000/rng_state_15.pth +3 -0
  36. checkpoint-9000/rng_state_2.pth +3 -0
  37. checkpoint-9000/rng_state_3.pth +3 -0
  38. checkpoint-9000/rng_state_4.pth +3 -0
  39. checkpoint-9000/rng_state_5.pth +3 -0
  40. checkpoint-9000/rng_state_6.pth +3 -0
  41. checkpoint-9000/rng_state_7.pth +3 -0
  42. checkpoint-9000/rng_state_8.pth +3 -0
  43. checkpoint-9000/rng_state_9.pth +3 -0
  44. checkpoint-9000/scheduler.pt +3 -0
  45. checkpoint-9000/trainer_state.json +0 -0
  46. checkpoint-9000/training_args.bin +3 -0
checkpoint-10000/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": false,
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-10000/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-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9c7ce53d43065076b2a07e2838e345b11a29f50475fc84d4ec9ff77dea99783
3
+ size 4538452374
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:301b67388c88847be9038733b4e3d40ac323064ca02d89cc94d2a7841327a8fd
3
+ size 6630551906
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4007efeeb3c5ddbca8edfad79e429ee2d0955f91c334fea70bfb9ba91a77405a
3
+ size 15984
checkpoint-10000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e2b0daa81f763fab72aeead183b2d6cff07ce87ab3b7f4e32192ef495076bfd
3
+ size 15984
checkpoint-10000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33f69188419118a0e5b694fa8c052192dac8430605f7d41c01f69bc1998848ae
3
+ size 15997
checkpoint-10000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a442736f152f171ff635532aeb13c9f6257aa6ad80ccd5b1b8f06aeb0d21b362
3
+ size 15997
checkpoint-10000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b770b2d6e00c0aa23e1e08d529e3421b8b6dca5033126af795ab15ff9c419383
3
+ size 15997
checkpoint-10000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19fb9c9588c1c9557e4a8b3571302d7bb53b85a16a161d142174c5826be4be1c
3
+ size 15997
checkpoint-10000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81a275eb353266cf218a261e2f6dd91f0474757578477c4e3e9504cea501c76a
3
+ size 15997
checkpoint-10000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66eed2a0c1b227cab29ed55c8f61b0fd75135016de30af7948f5778eddf59968
3
+ size 15997
checkpoint-10000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3271f4723e77bd4b137fe31b5b4826a4b0041ca421a760f800eede3444641833
3
+ size 15984
checkpoint-10000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f877aed60cc9c97e12bc9d3543cbbabf6bfc8967594482e3b9ec81463bc323e
3
+ size 15984
checkpoint-10000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:698cd1855f2af3f1200f2eaf4e472b5fae3bcc1fe8fd6a043a555e8fd1001506
3
+ size 15984
checkpoint-10000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70beff16b336f0ed96c3422893612a6443a4b98848b0f67316c930cc274116a1
3
+ size 15984
checkpoint-10000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24060b4467f3dbfef2aa5ab72895b89bc706a5fd9a9dee92bcf2df48e5efd058
3
+ size 15984
checkpoint-10000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ab847750c40c16514bdd34bf1b8e87abf4eb8381042073012068469015ee14c
3
+ size 15984
checkpoint-10000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b04ffbabd0ba22dcdb2d3a688760dd9adee77f966867bcba82c165ce8bed422
3
+ size 15984
checkpoint-10000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b66fcb4bde3a59e91a2cf0d605515915f3b2ba6b9772bf1f54e0a556ad50fd6e
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cecba43d7459ec5ce9f816c2e7f8c9a6c0974f0e469d7bde6bc33484cab04b5
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa580d00453a34ab8ec3cfdf9f330086ae3464ae148430385343a700076bbc15
3
+ size 5944
checkpoint-9000/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": false,
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-9000/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-9000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:668fcbd3957d62d64d36df8752111c4f150d380c9075eaadef5c2879f0da890c
3
+ size 4538452374
checkpoint-9000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac460505e4f8fb3b9ede0b130dd3d55cadafc7ada2b015a4899ea82451eb2d7f
3
+ size 6630537250
checkpoint-9000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c36f6b3f481f0e28474556427c65fa2d3cc9c35ba2e869bd60400c995f0b3de
3
+ size 15984
checkpoint-9000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:134ca0e113885db8242c898dbf2ed6cf476a4df79bc41a20427b567089d58afb
3
+ size 15984
checkpoint-9000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df4a74a0511c15d288d93401b6cc56f4778a99bd48bfa438538537bb1384ee4a
3
+ size 15997
checkpoint-9000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b49c56cd2f22e1c40e53f1673a1e973ab21c0cac05a95943cc8d47e2f71ed09
3
+ size 15997
checkpoint-9000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58eb5bbef68cf58e98c5fdb5bc975747167f617ed9bab2bb832ac418eb1ee294
3
+ size 15997
checkpoint-9000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c57948983e8a30f3012b74629dab1c4ff3b58a96722a898ad50099c28428ca90
3
+ size 15997
checkpoint-9000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52c123ac235a3a868bd96c26196c3dfed68e775fd703f485a6fb2d84c00ffd4f
3
+ size 15997
checkpoint-9000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cca4affe627518722ad6854ee51500ebb005f445c0f35381a8c93c30886fc683
3
+ size 15997
checkpoint-9000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8873ee5c9a0310352620e7b69052dc90b93d9f75201ea00b39f5e98f18244baa
3
+ size 15984
checkpoint-9000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9614bf47327a61ba44e94b511aa7d485fedfc2e88c1469e46fd64ee814e5808
3
+ size 15984
checkpoint-9000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4e02bfb0561644a8d0b1ed930cda0e4a6a4ea9d5264ef15901a660650338da1
3
+ size 15984
checkpoint-9000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb19b3acf8c22a4680b87f2aa904009a38edbc3db34375b87e9a225b6e49df3a
3
+ size 15984
checkpoint-9000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6568ccf76c66d960bd158f2e36e543aed5241eb30b5ae96abc83e99888cfe495
3
+ size 15984
checkpoint-9000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c67e9f776a738c19c4b65776a33fff5d1eeee6a53f1f6813664bb214db63950
3
+ size 15984
checkpoint-9000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c588a1b3087cf28a88606df367ea6dd26ac6065180f0ad1f06e0b273da88a5c8
3
+ size 15984
checkpoint-9000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f297344e1f74752f52893bde462ca82221e5b0cfa27bdf282873d62fbc6dae17
3
+ size 15984
checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e71621b7afdbc391d9e5784b68966c636091af8a60dceeb1c8486a8fbac2cd89
3
+ size 1064
checkpoint-9000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f03f56435a0fa7228f00c58cd5f56087f289afb034e722ad6d91f7b74e10699d
3
+ size 6008