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
15bfd39
·
verified ·
1 Parent(s): f02943c

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-27000/config.json +79 -0
  25. checkpoint-27000/generation_config.json +6 -0
  26. checkpoint-27000/model.safetensors +3 -0
  27. checkpoint-27000/optimizer.pt +3 -0
  28. checkpoint-27000/rng_state_0.pth +3 -0
  29. checkpoint-27000/rng_state_1.pth +3 -0
  30. checkpoint-27000/rng_state_10.pth +3 -0
  31. checkpoint-27000/rng_state_11.pth +3 -0
  32. checkpoint-27000/rng_state_12.pth +3 -0
  33. checkpoint-27000/rng_state_13.pth +3 -0
  34. checkpoint-27000/rng_state_14.pth +3 -0
  35. checkpoint-27000/rng_state_15.pth +3 -0
  36. checkpoint-27000/rng_state_2.pth +3 -0
  37. checkpoint-27000/rng_state_3.pth +3 -0
  38. checkpoint-27000/rng_state_4.pth +3 -0
  39. checkpoint-27000/rng_state_5.pth +3 -0
  40. checkpoint-27000/rng_state_6.pth +3 -0
  41. checkpoint-27000/rng_state_7.pth +3 -0
  42. checkpoint-27000/rng_state_8.pth +3 -0
  43. checkpoint-27000/rng_state_9.pth +3 -0
  44. checkpoint-27000/scheduler.pt +3 -0
  45. checkpoint-27000/trainer_state.json +0 -0
  46. checkpoint-27000/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": 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-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:d85853b04f5d38cddd757e7c59d905cb3a5dac03930d0505fea08429b33e75b4
3
+ size 4538452374
checkpoint-26000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f406b9d8677c6b7e7f2931d46a79b4f1950afca94384ed9460eb3c2c306abae
3
+ size 6630551906
checkpoint-26000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:060c9af8b14b70c329d97825dcddaf78af84f0899510a412ae4b2df6cd2e05f5
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 sha256:1e9d5d8296e236174d829d7b6537724d103d0c59e4ba072895e338a024fe8c3b
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 sha256:0ae0bb7b3a3c4dd44cc2d4dfae36a302b3d607fa6090fca92e77dd4cd31120ec
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 sha256:2122f76ad8bd6229384082f223dded5bfd70d66eae868bba4879a1271d91107a
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 sha256:26e4c72e9ba466399db4e978b72ea815f20b24bc486c365b51bcf4e945f3d0f2
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 sha256:550c7add58b31c7b28c52fb0db9f7bdab51c21a3f7c926e1499b8c71f7801ed1
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:fdfce0c7533aa21bd3842e51049ab7eef6b18181eb672dfda8dfeef1deae2fac
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 sha256:5f41a51f03cdc93104237256c4be091acd29008ed8ca6159ca40e605cb59f14b
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 sha256:74f78d131ce5917eefa84d0cdf244bf6dcf1a0cc6aadf9e11cb2b913d5942327
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:f51af414e15e2f32fbd33d61d47283ee2fa27a6c03a5b0f894b7134af7168193
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 sha256:9fe0a4299e3b773a5864a960d69eed53d5e3b128e18114d0cb5b85384c4d2965
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 sha256:867a6e8f743566256b0c145bf3fe27d49de863a57182e8deafe7e6f7c44f1bb3
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 sha256:50c60d7a82494350c7f19ddc47849e111ade9092f8dd92920492bf7f37262e41
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 sha256:2f2878c8b2f256e0f8feda4b050fb70bb1b0401eebb545fb58d008cdadad2699
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 sha256:16a7ee1594860304d1c033c32104ca394d08b7f4a98fe54d70cfb6b43753f76a
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 sha256:78ab6e358e887e2e55f4e3b70b9896da15ef7fcadf0edf637b2cea3e080015f3
3
+ size 15984
checkpoint-26000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4f68413d676a74d63c54337ce6dd2f8e89f3485d6493985914168fefb0d18bf
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 sha256:5f2fbd216cec84dde9c312e44a0eb00b00d2b32e08cdefb732e89ba43317197c
3
+ size 6008
checkpoint-27000/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-27000/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-27000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4df1e11169fa0d8f9bd45d1a023be788469d7136f27cc5519a2071912716c19
3
+ size 4538452374
checkpoint-27000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8442622ca20eb9e67e75747e70e05a5e7981767f707842e87b1c401775fd770f
3
+ size 6630551906
checkpoint-27000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aaf9f109018bb16f78696c58e730c6fbeab6abe605eae881103f2174b1b5866
3
+ size 15984
checkpoint-27000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:245abe1bae06edece57de2d4cc7be48a45a979bfd030530975411631aa5ad95d
3
+ size 15984
checkpoint-27000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96bca81f1a977d1793b423160e8cc39db304b1624d819388b29556e48dcaf8cb
3
+ size 15997
checkpoint-27000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4efdb4a133279271716e57bb131302a50d3feb1008f104f385b0ae8dffad476
3
+ size 15997
checkpoint-27000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5710057c0d061249bef2a4578917301e47cc82d8dbf2b9cc47fc799dba916f1
3
+ size 15997
checkpoint-27000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a735e12d683d434bcf7406edefb0b73deb5a271c6447658cedd72f7b3db8666
3
+ size 15997
checkpoint-27000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eaba001374349aed0ea3251d4db2472d23e0df24adb49b8f567ef678f1550226
3
+ size 15997
checkpoint-27000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a44ea449662b05ab551a4f9414b06d0c7421bb608fcfc6c01c4fb9bd6d32e44d
3
+ size 15997
checkpoint-27000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfa82ecaa62923fb9d15795b9430bf2a1be6c951c08df3b7f2265a031637468a
3
+ size 15984
checkpoint-27000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cedbe019408a27e86bb485769575bc9df570c7819f4fbb77d5e320fa9fe6d373
3
+ size 15984
checkpoint-27000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d1d15dee8ff77c697e374c9b929b068dd130f5fda6211741a98391b63632d4d
3
+ size 15984
checkpoint-27000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e33e9767b7546aaccfef528dd609268dd32171e3183141bc016adacd1d37560
3
+ size 15984
checkpoint-27000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bbd44792894a601a2e7d4ac90e8e8050a6240cab76ae2b10a4dfd9d6410daac
3
+ size 15984
checkpoint-27000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c9480075358faf98fb10ca2e1f649c75c5b54d4be280f9420e3855327dae84f
3
+ size 15984
checkpoint-27000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3185b60673d1a21f6e47a53c97d6f4d332202b5a51f6fa2d5d4bd226226c84f5
3
+ size 15984
checkpoint-27000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d8548ff7bf28c417e10235172521c7b0483a53bf4339497f81d9cccc0810654
3
+ size 15984
checkpoint-27000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc2439fe897da674625c81e482c94fb141c266ceacb253fec2439b944fbe9505
3
+ size 1064
checkpoint-27000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-27000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f68c3a788c758da15831710bfda6c391c09f299eddd7d5117255155a2a0e61e
3
+ size 5944