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
c27a9fe
·
verified ·
1 Parent(s): ffe4ae3

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-32000/config.json +79 -0
  2. checkpoint-32000/generation_config.json +6 -0
  3. checkpoint-32000/model.safetensors +3 -0
  4. checkpoint-32000/optimizer.pt +3 -0
  5. checkpoint-32000/rng_state_0.pth +3 -0
  6. checkpoint-32000/rng_state_1.pth +3 -0
  7. checkpoint-32000/rng_state_10.pth +3 -0
  8. checkpoint-32000/rng_state_11.pth +3 -0
  9. checkpoint-32000/rng_state_12.pth +3 -0
  10. checkpoint-32000/rng_state_13.pth +3 -0
  11. checkpoint-32000/rng_state_14.pth +3 -0
  12. checkpoint-32000/rng_state_15.pth +3 -0
  13. checkpoint-32000/rng_state_2.pth +3 -0
  14. checkpoint-32000/rng_state_3.pth +3 -0
  15. checkpoint-32000/rng_state_4.pth +3 -0
  16. checkpoint-32000/rng_state_5.pth +3 -0
  17. checkpoint-32000/rng_state_6.pth +3 -0
  18. checkpoint-32000/rng_state_7.pth +3 -0
  19. checkpoint-32000/rng_state_8.pth +3 -0
  20. checkpoint-32000/rng_state_9.pth +3 -0
  21. checkpoint-32000/scheduler.pt +3 -0
  22. checkpoint-32000/trainer_state.json +0 -0
  23. checkpoint-32000/training_args.bin +3 -0
checkpoint-32000/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-32000/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-32000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84975b065e0c033dc0133e8952629d31e133c57dd1557229e97b761321678d6e
3
+ size 4539487822
checkpoint-32000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a4c96b920102fa91a4824961d022bb744d6029415cde683b0221ca1cbc11644
3
+ size 6632345482
checkpoint-32000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8abea23026535fb402a76ccb0fbc04865eaa524f27f44cb8716fd6092258c18e
3
+ size 15984
checkpoint-32000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54943ecb2119dba9f9fa5e96c4f8f82e7033e82e200d240559ad08db5a5c86cb
3
+ size 15984
checkpoint-32000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff240b5744a1c1fe372bce2e850fae0c488c811711f161aa9d8b5ee40b3df8d4
3
+ size 15997
checkpoint-32000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:043f2b1a1b2a542cd3560d0b3d9f62944b8f7e9bb542924bdefd4af73afadb9c
3
+ size 15997
checkpoint-32000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ebebc46c4701cb9a84627efac43014cd7c579510bdb4cad5daeb922742f239d
3
+ size 15997
checkpoint-32000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aefcb7058690dad2999f09a847cb5df44399adfe564e8d393dac94218c8b01eb
3
+ size 15997
checkpoint-32000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63af56825822fb4576964c09b86a266de069b9c13aeb72caa501128f78d717e8
3
+ size 15997
checkpoint-32000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69b01e1c51dbe25b9738bfe2c9395b29006705a9862d0f741a60268422faff89
3
+ size 15997
checkpoint-32000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7e4844b43576016103b806a8253f46a24c9d33caeea4d20430a8e875575c961
3
+ size 15984
checkpoint-32000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d58067a8de18147f77cc801ff32b95d394b19d7656aacdde1d97dac303f700db
3
+ size 15984
checkpoint-32000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c278b014816586802fa79200032be42d58f71675fb10542a40736d954e8b7733
3
+ size 15984
checkpoint-32000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b92c3de4cb27195301c74f11b36bdb2180a1c8cfecd1ae03cbdefa7a154b232
3
+ size 15984
checkpoint-32000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f717a472f8a8927ea086e2fb6b380875eca1c0f6d8f2c139a8100a17d098e19
3
+ size 15984
checkpoint-32000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dde7cf5bddd39e420f0602273d38c81c6d045865c7dc7b27d320cc22046acd81
3
+ size 15984
checkpoint-32000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a848d046ca917f1e5a2adf6bad55ea28d35512aeb432e57cb05f389153d47c0b
3
+ size 15984
checkpoint-32000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d22b530bd92394435a4c1d1c15f6ecc52ffae74a5ad301fe4e154f3e0d6800d
3
+ size 15984
checkpoint-32000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e924d71d6576695678b978dceee56d6d28fcf33466daba824d5fcae75856609f
3
+ size 1064
checkpoint-32000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-32000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48b293265efce8a2e927ff2650b5151d04c1112f3d313b20f60e5628e3694f52
3
+ size 6008