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
181bd4a
·
verified ·
1 Parent(s): fd6c5a0

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-18000/config.json +79 -0
  2. checkpoint-18000/generation_config.json +6 -0
  3. checkpoint-18000/model.safetensors +3 -0
  4. checkpoint-18000/optimizer.pt +3 -0
  5. checkpoint-18000/rng_state_0.pth +3 -0
  6. checkpoint-18000/rng_state_1.pth +3 -0
  7. checkpoint-18000/rng_state_10.pth +3 -0
  8. checkpoint-18000/rng_state_11.pth +3 -0
  9. checkpoint-18000/rng_state_12.pth +3 -0
  10. checkpoint-18000/rng_state_13.pth +3 -0
  11. checkpoint-18000/rng_state_14.pth +3 -0
  12. checkpoint-18000/rng_state_15.pth +3 -0
  13. checkpoint-18000/rng_state_2.pth +3 -0
  14. checkpoint-18000/rng_state_3.pth +3 -0
  15. checkpoint-18000/rng_state_4.pth +3 -0
  16. checkpoint-18000/rng_state_5.pth +3 -0
  17. checkpoint-18000/rng_state_6.pth +3 -0
  18. checkpoint-18000/rng_state_7.pth +3 -0
  19. checkpoint-18000/rng_state_8.pth +3 -0
  20. checkpoint-18000/rng_state_9.pth +3 -0
  21. checkpoint-18000/scheduler.pt +3 -0
  22. checkpoint-18000/trainer_state.json +0 -0
  23. checkpoint-18000/training_args.bin +3 -0
  24. checkpoint-19000/config.json +79 -0
  25. checkpoint-19000/generation_config.json +6 -0
  26. checkpoint-19000/model.safetensors +3 -0
  27. checkpoint-19000/optimizer.pt +3 -0
  28. checkpoint-19000/rng_state_0.pth +3 -0
  29. checkpoint-19000/rng_state_1.pth +3 -0
  30. checkpoint-19000/rng_state_10.pth +3 -0
  31. checkpoint-19000/rng_state_11.pth +3 -0
  32. checkpoint-19000/rng_state_12.pth +3 -0
  33. checkpoint-19000/rng_state_13.pth +3 -0
  34. checkpoint-19000/rng_state_14.pth +3 -0
  35. checkpoint-19000/rng_state_15.pth +3 -0
  36. checkpoint-19000/rng_state_2.pth +3 -0
  37. checkpoint-19000/rng_state_3.pth +3 -0
  38. checkpoint-19000/rng_state_4.pth +3 -0
  39. checkpoint-19000/rng_state_5.pth +3 -0
  40. checkpoint-19000/rng_state_6.pth +3 -0
  41. checkpoint-19000/rng_state_7.pth +3 -0
  42. checkpoint-19000/rng_state_8.pth +3 -0
  43. checkpoint-19000/rng_state_9.pth +3 -0
  44. checkpoint-19000/scheduler.pt +3 -0
  45. checkpoint-19000/trainer_state.json +0 -0
  46. checkpoint-19000/training_args.bin +3 -0
checkpoint-18000/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-18000/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-18000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b691963e9b4fec41c275f7b2a3caf3aaf846b831174e8873d62558cc3c42689c
3
+ size 4538452374
checkpoint-18000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e01ff1eca457e089baa52b362e95ed442b009185c49824cea2772bafa53b1614
3
+ size 6630551906
checkpoint-18000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aef8d0bf7d622ee7956c8d923d269e7239aa7edc544b99e2d7648a715a2621a6
3
+ size 15984
checkpoint-18000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb9dbe70c16d45a8843685f5c0c95ae4d264854b5dc002c8a18b9bb76d712c00
3
+ size 15984
checkpoint-18000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4883aa9288001fbaa36b3e8dd70556ae7d92fc3f82de593828922cf741e83f68
3
+ size 15997
checkpoint-18000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ce15787ab332a3a2e7a530004426b0f87593afe9a26b60a65a219567d90686f
3
+ size 15997
checkpoint-18000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6448e7da7a475caf35270ec5700724e7e6bd7d4cd95ed2e8ae78e541f624055a
3
+ size 15997
checkpoint-18000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bec15bb3e7c50d892192e164bd3298c3288e7319ee010c2aad1e3b2dfebc43a
3
+ size 15997
checkpoint-18000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a48b78575952aee626e207579deb7a746712d379d2cea06d13e031d79346f0f
3
+ size 15997
checkpoint-18000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d8333073bd2ce945db2bf320aea87bba81e1315d68a20999de8de5ba3914b44
3
+ size 15997
checkpoint-18000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a605a7b9fff90ec1f1065cf47e3086cb690d92f659b44c7c44bb2d62315c2bea
3
+ size 15984
checkpoint-18000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7efa0fa0109d80c06705f0f41e1aca93406919d4e6cada923a0b852a0076ee7c
3
+ size 15984
checkpoint-18000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:195f9b81e50d134c4e2480159c966df022264dfdbbd3b74d2978a7e3c336f76f
3
+ size 15984
checkpoint-18000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:825580a7c8623db7fc223c988750aa3e948e1253e6badcaa6ea4be9a6635d64e
3
+ size 15984
checkpoint-18000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:767aac295f6d4a01c3f9ac77ba6a71b6aaad1e6f1c015544c855573c88366e58
3
+ size 15984
checkpoint-18000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:508e2d6cf1f3577ecbc8782811a61f04ebdf9573df296a183d5acd9c363e4058
3
+ size 15984
checkpoint-18000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6936a92c1ce7eb436cddea76ac5bb85d0ef99b1113d2bc5cb918f1bb45d4a77
3
+ size 15984
checkpoint-18000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3112016101c47bfa4d23e48613afe1d087b5f4cb38d2740803f38b98194d1be
3
+ size 15984
checkpoint-18000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb31aaff61ae7eeb7d8832032e6a4a21583cf7098d242c851d0905f05fd82883
3
+ size 1064
checkpoint-18000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-18000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a03243a26f32351b899b679a089310323b308d7fe1f93f81ead3d6e1cc0d5d0b
3
+ size 6008
checkpoint-19000/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-19000/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-19000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eca59cb6fb58c88fd2618e694943297fa963e1452b7d8982ea77bd3da3545e82
3
+ size 4538452374
checkpoint-19000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32e2e839262f649e725727f743924f9fdbb979c7859046e3e982b68d0078cb13
3
+ size 6630551906
checkpoint-19000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02102152fb43a818b4846ee18f72f974f667cd3f177fee0aacdfa1757a8d42e1
3
+ size 15984
checkpoint-19000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4bcf2941623956b4db3c25c5e553da4ddff75bfb625bbc9ce9597ec2a0f0839e
3
+ size 15984
checkpoint-19000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:646d3a8a6ca1505ded2a93294c226a68d61d4ad10d903057f1e059dfc81d1c14
3
+ size 15997
checkpoint-19000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82c73fd42c83b2cb3418c9a2e3abec375e5459d8d17d2e7d4eb2f83a3df53104
3
+ size 15997
checkpoint-19000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f87c9e006afa0f8c50102a2f7d5cea4e881499162e15f2f1f3942d28f5fd32e0
3
+ size 15997
checkpoint-19000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3537e32ec8e893fe98a8dd51ae32c36450aa0e2c134d2cfcdaec748da325dd5
3
+ size 15997
checkpoint-19000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86702c4a3a8ec53715910c9fc14f5832e04299dd774473247a521d26193d52b1
3
+ size 15997
checkpoint-19000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:caba80c43509901f4323b873ede0262d63ff471ffc70220fd74363d0c66b3a60
3
+ size 15997
checkpoint-19000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45e67e4ecba800cb4452a79af5cc77fd2678082717dc33865450f71e79b3492a
3
+ size 15984
checkpoint-19000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af38f5b70b6b831e236e75b464ef073994d5a0aa3b1644959c0807eb0ffaa36b
3
+ size 15984
checkpoint-19000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9bd38b6b88ae7cc1491c596f7804ca33ce2cc21f0ab5a25a5dc7be2683a81922
3
+ size 15984
checkpoint-19000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b16a8bdb98e7b1612c6af8133abecb4249df868be786b4fe7f32183465010b1c
3
+ size 15984
checkpoint-19000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df9c5037d77ba5597efdb72784662926719ccf2a74cde6695672c327e2f7d8a9
3
+ size 15984
checkpoint-19000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9adc780bebc79afbafc5555a4b6a920b14f1a3196f305fdab1b18f0f0b02dcb
3
+ size 15984
checkpoint-19000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61554aa1035f5d4a690224128cbc9a83932ee5060a6bba588695b47a85856a5b
3
+ size 15984
checkpoint-19000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fddfdb01467416cce7adf33e3b6dc2eaaebeae2cc97718cb53f4f64505e468d7
3
+ size 15984
checkpoint-19000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d85aa21a9c83245966ad19e50575c1e6603ce6de05df3f08a420cfcfca806271
3
+ size 1064
checkpoint-19000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-19000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a18ca3a56d89f67da6a781c62ec32c37596ffb5e212cbdb6e88c78942620f77d
3
+ size 5944