Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
a6cb87f
·
verified ·
1 Parent(s): 46fb8c8

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-26000/config.json +79 -0
  2. checkpoint-26000/generation_config.json +6 -0
  3. checkpoint-26000/model.safetensors +3 -0
  4. checkpoint-26000/optimizer.pt +3 -0
  5. checkpoint-26000/rng_state_0.pth +3 -0
  6. checkpoint-26000/rng_state_1.pth +3 -0
  7. checkpoint-26000/rng_state_10.pth +3 -0
  8. checkpoint-26000/rng_state_11.pth +3 -0
  9. checkpoint-26000/rng_state_12.pth +3 -0
  10. checkpoint-26000/rng_state_13.pth +3 -0
  11. checkpoint-26000/rng_state_14.pth +3 -0
  12. checkpoint-26000/rng_state_15.pth +3 -0
  13. checkpoint-26000/rng_state_2.pth +3 -0
  14. checkpoint-26000/rng_state_3.pth +3 -0
  15. checkpoint-26000/rng_state_4.pth +3 -0
  16. checkpoint-26000/rng_state_5.pth +3 -0
  17. checkpoint-26000/rng_state_6.pth +3 -0
  18. checkpoint-26000/rng_state_7.pth +3 -0
  19. checkpoint-26000/rng_state_8.pth +3 -0
  20. checkpoint-26000/rng_state_9.pth +3 -0
  21. checkpoint-26000/scheduler.pt +3 -0
  22. checkpoint-26000/trainer_state.json +0 -0
  23. checkpoint-26000/training_args.bin +3 -0
  24. checkpoint-28000/config.json +79 -0
  25. checkpoint-28000/generation_config.json +6 -0
  26. checkpoint-28000/model.safetensors +3 -0
  27. checkpoint-28000/optimizer.pt +3 -0
  28. checkpoint-28000/rng_state_0.pth +3 -0
  29. checkpoint-28000/rng_state_1.pth +3 -0
  30. checkpoint-28000/rng_state_10.pth +3 -0
  31. checkpoint-28000/rng_state_11.pth +3 -0
  32. checkpoint-28000/rng_state_12.pth +3 -0
  33. checkpoint-28000/rng_state_13.pth +3 -0
  34. checkpoint-28000/rng_state_14.pth +3 -0
  35. checkpoint-28000/rng_state_15.pth +3 -0
  36. checkpoint-28000/rng_state_2.pth +3 -0
  37. checkpoint-28000/rng_state_3.pth +3 -0
  38. checkpoint-28000/rng_state_4.pth +3 -0
  39. checkpoint-28000/rng_state_5.pth +3 -0
  40. checkpoint-28000/rng_state_6.pth +3 -0
  41. checkpoint-28000/rng_state_7.pth +3 -0
  42. checkpoint-28000/rng_state_8.pth +3 -0
  43. checkpoint-28000/rng_state_9.pth +3 -0
  44. checkpoint-28000/scheduler.pt +3 -0
  45. checkpoint-28000/trainer_state.json +0 -0
  46. checkpoint-28000/training_args.bin +3 -0
checkpoint-26000/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-26000/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-26000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:608659560f31aaedcb9046e1e9c0b04bf5549bfd33d4ea70c0c14ed9c1aac0f8
3
+ size 4539487822
checkpoint-26000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6d5387113b9c4b573f40948a9c69c0a25dd8097ef76cb36c1e8d1fadfa7447c
3
+ size 6632345482
checkpoint-26000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8162686f24f98896f7f524b5f9cdf216d3b181246fad9964f7547bd0e579c01b
3
+ size 15984
checkpoint-26000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0be662bdb54b76eebede631d5cf4232a5f6ac8c97687b91405e3295d2a7721ee
3
+ size 15984
checkpoint-26000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64e4c6ed1158ec2030604eec894a6b46d32d0b5abd04c055df0384d927ead3cc
3
+ size 15997
checkpoint-26000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97498485b38d210d8f7ddf64237d1fe708057c190ebc1f740f1d3b1399c24b30
3
+ size 15997
checkpoint-26000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2993fd461107582a5557734ce160eb54562d49099e41fba546f52087293110b
3
+ size 15997
checkpoint-26000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f517d0712731b3ecb91fa261438e711a1c80dd8a304fb7726608db957e5b6677
3
+ size 15997
checkpoint-26000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64d180793f1f1622c63dcb83842e32f9ca76c46950278887cfa26c2db4e250e3
3
+ size 15997
checkpoint-26000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:133b9743ffa9dd0f83701ca074dca1b3ac3063f1821d00d9b5952e7204454b65
3
+ size 15997
checkpoint-26000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f89e1170e86abc78f87691635795cee14c6c1178d29b327744af9183eed55bd3
3
+ size 15984
checkpoint-26000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52da5b5d2a6b324d722e6e088ae66772d97b419e13d61edfb06469266a72a126
3
+ size 15984
checkpoint-26000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a88c0c4ce43a046b28d4ab94deab97e5aab02447a9cf0462bc4ee5f5b99b4a9
3
+ size 15984
checkpoint-26000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:283148d63752b5c3ae73ac79547a84b08a2a3b3d835862684b0ffcbcbb6c629b
3
+ size 15984
checkpoint-26000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97264a5efda301a514c33a778e50b1dd38434562602e2afdc8832a308eba8fb7
3
+ size 15984
checkpoint-26000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db4028d576b0953868e30c4b9ce41ab68b0e71913c7296de5801ba7ba683d725
3
+ size 15984
checkpoint-26000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bab9b1047e6ec655b705689822e0c2fcefee54e9f4425c3534b07fe962e65133
3
+ size 15984
checkpoint-26000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:804d632792dfd8719e606e0f1dfdab1dea8f55907879609ac36d1adc5761d55a
3
+ size 15984
checkpoint-26000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb5bd17159554b6bae0c60fdfe64e0dcf2d23c92fbd079eda3dc0030b25f44a7
3
+ size 1064
checkpoint-26000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-26000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f5124197eb9fb4d98668a93520441aea27beac1c7544177e2a4ecb7c17ee234
3
+ size 6008
checkpoint-28000/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-28000/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-28000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73f5788db33bd0d8db7733cdecfb5597e85ea32673eeac9143f022d09c6a7619
3
+ size 4539487822
checkpoint-28000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4ce008306ce213b66cd0b150b2996bfd7ca25e016a568e785255e42f23754aa
3
+ size 6632345482
checkpoint-28000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d554d5bc9e2264d1fac7f9681aac393ccab36523204f39d45680b14741be8ac1
3
+ size 15984
checkpoint-28000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd52a1b2ffba2fabddcdcdce52c25c9556181ec05e59297a927e8306250793aa
3
+ size 15984
checkpoint-28000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7b8a945650a455bcdd654fe59b269abef5d7c1f829c504f167c4f56cb4c0944
3
+ size 15997
checkpoint-28000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1405a72309227ffb7cb6a3392dc8d862a3b3bf3208f1d125d60009ae28373178
3
+ size 15997
checkpoint-28000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ced211ba83800485c79e997bf855c94d6d30756ea7a4a0abc24f6b88aa6b88f6
3
+ size 15997
checkpoint-28000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb294cb38f005614e2703211574fda5d2c81bd9c84bb1083b9b6ce573d5b0cb3
3
+ size 15997
checkpoint-28000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf2f5f9db6f6f79854470b87b86467fb3af24f8836e989d709823d2f977d1a67
3
+ size 15997
checkpoint-28000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fc6cc898124ba6aeb5946b718a3ab4f37e1b9696e5decaf383cc5072e2dd408
3
+ size 15997
checkpoint-28000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:976c18dfa1ac6c3a0490bfa245d48aec4e1d7645a9bda437112bc0368b48527d
3
+ size 15984
checkpoint-28000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ee44720b8d3940962ae12f36c05414ee5c06c1253f3381a21ad52e551a32752
3
+ size 15984
checkpoint-28000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9992b28a878bc4ab46f86a3b6bbb08578260c038ba767b36f57427df217a6064
3
+ size 15984
checkpoint-28000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4d3749d570440687302ffa5f8a0d1c24bfe2c779f8e395c65b200f558a7af85
3
+ size 15984
checkpoint-28000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:178dbcb1c501518e389938754964c1382c6df8bd0bcf0daa1fb8bfcff8d71711
3
+ size 15984
checkpoint-28000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c38595aad389592d65de2a8bb7272ada2b962636d23574f6d60c2974b05e1bfc
3
+ size 15984
checkpoint-28000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d09903a602406829f1df78b8004ecf8ec104537b190b1c353007aefe9c8c7e3a
3
+ size 15984
checkpoint-28000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fca14a7b665d0528caee12d911f5daacf8f1bb8e881704eede251d4e94590221
3
+ size 15984
checkpoint-28000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aefb40a68c60d352d51b0614becdc58a87c023271272a5fb12f0deca8b90e6a
3
+ size 1064
checkpoint-28000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-28000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:330b75d32aad6fa435bfc84240eda106d24a4cf1f660eaee558f2672906d4409
3
+ size 6008