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
3e70f14
·
verified ·
1 Parent(s): d78e561

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-59000/config.json +79 -0
  2. checkpoint-59000/generation_config.json +6 -0
  3. checkpoint-59000/model.safetensors +3 -0
  4. checkpoint-59000/optimizer.pt +3 -0
  5. checkpoint-59000/rng_state_0.pth +3 -0
  6. checkpoint-59000/rng_state_1.pth +3 -0
  7. checkpoint-59000/rng_state_10.pth +3 -0
  8. checkpoint-59000/rng_state_11.pth +3 -0
  9. checkpoint-59000/rng_state_12.pth +3 -0
  10. checkpoint-59000/rng_state_13.pth +3 -0
  11. checkpoint-59000/rng_state_14.pth +3 -0
  12. checkpoint-59000/rng_state_15.pth +3 -0
  13. checkpoint-59000/rng_state_2.pth +3 -0
  14. checkpoint-59000/rng_state_3.pth +3 -0
  15. checkpoint-59000/rng_state_4.pth +3 -0
  16. checkpoint-59000/rng_state_5.pth +3 -0
  17. checkpoint-59000/rng_state_6.pth +3 -0
  18. checkpoint-59000/rng_state_7.pth +3 -0
  19. checkpoint-59000/rng_state_8.pth +3 -0
  20. checkpoint-59000/rng_state_9.pth +3 -0
  21. checkpoint-59000/scheduler.pt +3 -0
  22. checkpoint-59000/trainer_state.json +0 -0
  23. checkpoint-59000/training_args.bin +3 -0
  24. checkpoint-60000/config.json +79 -0
  25. checkpoint-60000/generation_config.json +6 -0
  26. checkpoint-60000/model.safetensors +3 -0
  27. checkpoint-60000/optimizer.pt +3 -0
  28. checkpoint-60000/rng_state_0.pth +3 -0
  29. checkpoint-60000/rng_state_1.pth +3 -0
  30. checkpoint-60000/rng_state_10.pth +3 -0
  31. checkpoint-60000/rng_state_11.pth +3 -0
  32. checkpoint-60000/rng_state_12.pth +3 -0
  33. checkpoint-60000/rng_state_13.pth +3 -0
  34. checkpoint-60000/rng_state_14.pth +3 -0
  35. checkpoint-60000/rng_state_15.pth +3 -0
  36. checkpoint-60000/rng_state_2.pth +3 -0
  37. checkpoint-60000/rng_state_3.pth +3 -0
  38. checkpoint-60000/rng_state_4.pth +3 -0
  39. checkpoint-60000/rng_state_5.pth +3 -0
  40. checkpoint-60000/rng_state_6.pth +3 -0
  41. checkpoint-60000/rng_state_7.pth +3 -0
  42. checkpoint-60000/rng_state_8.pth +3 -0
  43. checkpoint-60000/rng_state_9.pth +3 -0
  44. checkpoint-60000/scheduler.pt +3 -0
  45. checkpoint-60000/trainer_state.json +0 -0
  46. checkpoint-60000/training_args.bin +3 -0
checkpoint-59000/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-59000/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-59000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1b989855e6f7d07dc234d907eca5cbfb44c2c144092467dba13e141fa4c91aa
3
+ size 4539487822
checkpoint-59000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc88db7798a001c13357bdc755e72b8773d4e406ee009f96f0c8b994b844cefc
3
+ size 6632345482
checkpoint-59000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0956aeb33632e1995e6512d942229bf8ad811890c6cbbfd9f6c89a26789b199c
3
+ size 15984
checkpoint-59000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:860c69639a95ce1f8011e342b1ce668b3242a7117cf0cfc6ad41925517161fa2
3
+ size 15984
checkpoint-59000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5df929454f2be7184d6ba29cd43ae1728e2f8da46a518da24b8fcd755f4317ec
3
+ size 15997
checkpoint-59000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06d13c767bb7a5dcba7351463e9e8d851d11174bed77f4a081c2daa8994296ee
3
+ size 15997
checkpoint-59000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:768613bff62cfa6df967c6240fae2b78083dfcd02e9f8bb1d21ded294984402f
3
+ size 15997
checkpoint-59000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf241bb93501373aa149061c6e95473ae7e92bbfac5639e2aa630beeba6c50d0
3
+ size 15997
checkpoint-59000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39ce73c3740b7c54026c88acd319b711bb70f82cc0d657cf59d85b75a3bb3975
3
+ size 15997
checkpoint-59000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:557bd5fb97354e70457256143ff852a7a3c03b4e131674ffb26ec1ce14a5cfff
3
+ size 15997
checkpoint-59000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d21872f9b2bdb752b110d93e0ebceb0547c2d05c42b174b460f41dc53f7d1df6
3
+ size 15984
checkpoint-59000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11b61f45dd4edb845b869688bb10043505f3b6263373ab8cc835e9cc65d0973c
3
+ size 15984
checkpoint-59000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:462278775d77e7e161f8d76558097a95da70a2ceb572bfe1ff4ce1f22de12bc9
3
+ size 15984
checkpoint-59000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a431fd1c72d6efe460ce29357a72e9377b0a072190c8a9988926a5b5f9cfd66
3
+ size 15984
checkpoint-59000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90300620feca822ecd44bbc22b669ba30ef58519001ebc46b66542b6c8b4bb52
3
+ size 15984
checkpoint-59000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394776c537b60257b1cd15cbff67625af7a7ea193334aee2c46f93a11458f644
3
+ size 15984
checkpoint-59000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b5819cc6fcec0642aa4996f37e4c37ce2da598481efcddc68e3d9332b234ef9
3
+ size 15984
checkpoint-59000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e3393a52ba6ad28a9dbbd8a9409f204b108c86c58d4aec16a7a22dc418a3e5c
3
+ size 15984
checkpoint-59000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83b6b2c6d81d7c05a9db49df9bcc4349862c8baa05d544649f709611ffa15bab
3
+ size 1064
checkpoint-59000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-59000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1dd2624bda912ce1bba5a474333c5296f656e129e9c368ddfc3510efcef9b3cc
3
+ size 6008
checkpoint-60000/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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:359b8cbafa93ce9054cba5a3608e064a1ff3d4ccbac749827f9f723d3c757d28
3
+ size 4539487822
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f31e4223ebf2a0fd103f7cb2b64a590bc102bd0d227b7e0a25015502160f282
3
+ size 6632345482
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f39b879521d2cac3c683478491346772ef13e4baf4af1b1530f96186f92c333
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e32d81511701029f3ac480240e68511cbfae40862761722df9601d3ff6303996
3
+ size 15984
checkpoint-60000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eee15847b3c25037b95631194dc7bfe8bbf44bfe9b462e318cc7ef216ca2557f
3
+ size 15997
checkpoint-60000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:610ad990479822add796ed7d1184e4136d3aebedf176916e53b5b4c9754e8790
3
+ size 15997
checkpoint-60000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d04d89dbfe36af27e806453001d22193f219d94e74e74cf00b57d0b2643fc684
3
+ size 15997
checkpoint-60000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e3ef3e68f3ac56e25967d16e8c065d415b2a97322777dd27a8c38ece77e0b5b
3
+ size 15997
checkpoint-60000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0e29975dea6f3970f53ffa47546d12460c347107c98653566097846017f77be
3
+ size 15997
checkpoint-60000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d4392a183208b35e413cc8b2671f081ab143dfe5eb880a78ed40fe356dd2531
3
+ size 15997
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8b262cd54d9888efcb829732116987c407fc7e01955722b4e301a10cfd76a79
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b88278dbed38f5dfdf21f8ba51e969834cd444bb88f4b3ab7f6fa648a3fd689
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:033cf0e07529fd01399208a29610203b323aa4707adeb353f4d578024ed1f477
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:981c896c7845635a91bae5a6eacf11bd61605ec2a42a0bd4b6d15458069b439a
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9c73372134e14892d5644c763589108492d01d7a60beb39f77ed5f40a6dc540
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0df7f5cbe019a8f091df4826186d93cf155f64903e3c0e6f809d1ae3f17eb10d
3
+ size 15984
checkpoint-60000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e91506c2201cb8ceb1cc702bd70e96dfc5b634da0ebaf25f4fc155254a4c334
3
+ size 15984
checkpoint-60000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8c6ccb46077d624f8e8ebf012af37805cf99e00c99d888a96a463551cc7e1ef
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d97e514637e434b2bf16206f070fa73cdd1886ef533977b76474b9e329b4e4c
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1dd2624bda912ce1bba5a474333c5296f656e129e9c368ddfc3510efcef9b3cc
3
+ size 6008