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
f02943c
·
verified ·
1 Parent(s): 181bd4a

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-20000/config.json +79 -0
  2. checkpoint-20000/generation_config.json +6 -0
  3. checkpoint-20000/model.safetensors +3 -0
  4. checkpoint-20000/optimizer.pt +3 -0
  5. checkpoint-20000/rng_state_0.pth +3 -0
  6. checkpoint-20000/rng_state_1.pth +3 -0
  7. checkpoint-20000/rng_state_10.pth +3 -0
  8. checkpoint-20000/rng_state_11.pth +3 -0
  9. checkpoint-20000/rng_state_12.pth +3 -0
  10. checkpoint-20000/rng_state_13.pth +3 -0
  11. checkpoint-20000/rng_state_14.pth +3 -0
  12. checkpoint-20000/rng_state_15.pth +3 -0
  13. checkpoint-20000/rng_state_2.pth +3 -0
  14. checkpoint-20000/rng_state_3.pth +3 -0
  15. checkpoint-20000/rng_state_4.pth +3 -0
  16. checkpoint-20000/rng_state_5.pth +3 -0
  17. checkpoint-20000/rng_state_6.pth +3 -0
  18. checkpoint-20000/rng_state_7.pth +3 -0
  19. checkpoint-20000/rng_state_8.pth +3 -0
  20. checkpoint-20000/rng_state_9.pth +3 -0
  21. checkpoint-20000/scheduler.pt +3 -0
  22. checkpoint-20000/trainer_state.json +0 -0
  23. checkpoint-20000/training_args.bin +3 -0
checkpoint-20000/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-20000/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-20000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6988a4b8fecc8fa10419726fe3095057caae928f1345740c545d8464949382de
3
+ size 4538452374
checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:798ae96a9db254d55e4aa25afe4d425a28499ab1b3909789d8e97f01f44df83c
3
+ size 6630551906
checkpoint-20000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98b5bcc78e4dd8a5ffeedac63c1026dda2d87fffb26f3689f80792623f795393
3
+ size 15984
checkpoint-20000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8b39d63ac043a1b4d5f47442adca5f7ed723e62b10e631e1f07ee1fdf6eef35
3
+ size 15984
checkpoint-20000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18e9cc7af567c78580d9dece31071634b7b76e51730d1cbbb803907743ff633f
3
+ size 15997
checkpoint-20000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b95981bf418f77919e6b6532804ce3945e7f83ddc153fb5ab1206654836fa1cb
3
+ size 15997
checkpoint-20000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:920dd743fd35875c121e4a7a4460fdd1d8cdd190f2d656ee67e830c998f3788f
3
+ size 15997
checkpoint-20000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac14aed8ec10a465564afccecff9ace96fa83dd506b8069fdd44c850fbf1a05c
3
+ size 15997
checkpoint-20000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1460e5544c32bd5e25ea5bfed028f27fcd6c24ea54b907eff05a2229e8f5bd80
3
+ size 15997
checkpoint-20000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac643e7dbff7fe251628a7dfb3a2d261eb5364bbd10355bf792355477249ef39
3
+ size 15997
checkpoint-20000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e52f54e6279133d029c5b4c0ba2f6cac861a5fe2d114f4ac1a73c55b9ae4677e
3
+ size 15984
checkpoint-20000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49fc214d40daa32a063f57aac2d841fe926ee82ed3e20acdf2112697684d838d
3
+ size 15984
checkpoint-20000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a149257de142b0a3b746cca740a03d901939103123e91f06a419c3ca3aede6e7
3
+ size 15984
checkpoint-20000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f44134a35341cf53c791efa2acae36cf1f0c050c9d20e91ba0a7a3f08825a89
3
+ size 15984
checkpoint-20000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f2716b80c190f17ba4a8f5e8c36e8fc11809ab07290e6cc2869fd601686a7b6
3
+ size 15984
checkpoint-20000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20b52eb0a590c6d77a141a3f44c8e3098ff82f6007610002faedf417310a0de6
3
+ size 15984
checkpoint-20000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f2993a80c2491b1ee2f1d1ace00164ee6b69e3bf68715e27e27a89afda3a777
3
+ size 15984
checkpoint-20000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1826b10c3fa87ff1f080154c93f24ddcdc9d747f565a96f583fe6cc54691d465
3
+ size 15984
checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c84aebd00fa56a3c7d8e7c3e150905f18061ed98c1a033c0c777cbf63241f81
3
+ size 1064
checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0f072e0fe2062cb44b87124af7501da064ab58c96f5542a014b1d2c15e3037a
3
+ size 6008