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
0f49ab6
·
verified ·
1 Parent(s): 83ab8c8

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-124000/config.json +79 -0
  3. checkpoint-124000/generation_config.json +6 -0
  4. checkpoint-124000/model.safetensors +3 -0
  5. checkpoint-124000/optimizer.pt +3 -0
  6. checkpoint-124000/rng_state_0.pth +3 -0
  7. checkpoint-124000/rng_state_1.pth +3 -0
  8. checkpoint-124000/rng_state_10.pth +3 -0
  9. checkpoint-124000/rng_state_11.pth +3 -0
  10. checkpoint-124000/rng_state_12.pth +3 -0
  11. checkpoint-124000/rng_state_13.pth +3 -0
  12. checkpoint-124000/rng_state_14.pth +3 -0
  13. checkpoint-124000/rng_state_15.pth +3 -0
  14. checkpoint-124000/rng_state_2.pth +3 -0
  15. checkpoint-124000/rng_state_3.pth +3 -0
  16. checkpoint-124000/rng_state_4.pth +3 -0
  17. checkpoint-124000/rng_state_5.pth +3 -0
  18. checkpoint-124000/rng_state_6.pth +3 -0
  19. checkpoint-124000/rng_state_7.pth +3 -0
  20. checkpoint-124000/rng_state_8.pth +3 -0
  21. checkpoint-124000/rng_state_9.pth +3 -0
  22. checkpoint-124000/scheduler.pt +3 -0
  23. checkpoint-124000/trainer_state.json +3 -0
  24. checkpoint-124000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -60,3 +60,4 @@ checkpoint-119000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
60
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-122000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-123000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
60
  checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-122000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-123000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
+ checkpoint-124000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-124000/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-124000/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-124000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89abd53d9578cb0cfa89ad3e79c8a3c9bfe3b98944455056d187277c418daf6f
3
+ size 4538452374
checkpoint-124000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd8b1afc14b05af9e67622dc442785e9cbe065932fc96a60ad3b06f4bebb2797
3
+ size 6630551906
checkpoint-124000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57a5148242227f7c8c623cdee498468ebd58633d2ef03f191912f01999f2896e
3
+ size 15984
checkpoint-124000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f4039b488ca367351d0170da4f932e3bcab51dd7d494fde00718e67b11303dc
3
+ size 15984
checkpoint-124000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:265911d0c8fae7e0a59dc5af31b2a2867fd656c4f95609951905d987c53d6c06
3
+ size 15997
checkpoint-124000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7285b5f27ccdceb3751047294b4f558c9ffe065ce9ef5df2ab9b5e273f5c865
3
+ size 15997
checkpoint-124000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a89881762656ce6b19be125848fd1e9623f54fbb241ab48bc28907688ef2becb
3
+ size 15997
checkpoint-124000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fff7e80f3628099578955399d12c5d491403083ac1c21a8140e7433853f55b38
3
+ size 15997
checkpoint-124000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27446d632632faa17be9fc25691270883faf4b07d09605d72a14d792a8a480b3
3
+ size 15997
checkpoint-124000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04a4daf4255e54d29ee0e538361570018483c04c2794e9728d4450d6de86281b
3
+ size 15997
checkpoint-124000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e02a5c4184a23cecbf44edda17822043b5f1598cf73a6e6562c0a17df4852587
3
+ size 15984
checkpoint-124000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a592faae782210ce71f976261f810f8b0fb979780904b756aed9da6e71c09e29
3
+ size 15984
checkpoint-124000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ce14b1203d053f2abc037f953364b0f05f51610450d3065d8c093f225f9fe2e
3
+ size 15984
checkpoint-124000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:247fe36a6ec8b2618b8d9e1a9ad8386856e28e287ed4913838529948a7af125e
3
+ size 15984
checkpoint-124000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26e709d10a47bc0dd951fa1593646ab9291d3d3b35140ec6d6c3ba3c46c42086
3
+ size 15984
checkpoint-124000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3038eb8f610f8518d5db7b4c9a0a056a0a9d5910c035ecb6811f46e953905c2d
3
+ size 15984
checkpoint-124000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ab1ddca1a4cac866d3813fdd6c9de89ebfafaf57cdf07c922eb888bb773a6b4
3
+ size 15984
checkpoint-124000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a4f6a401bd52afbd25c172aedf4c9b63c11ca544acf93faf1596e96d863e95b
3
+ size 15984
checkpoint-124000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51a131656a7a8b4a42e4af20bbf818c54d46624ad54959fe96934ce337c9c533
3
+ size 1064
checkpoint-124000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:970c410e80567b4138e71cc4b09d4f648b083f4bc833404eb437c8aadd55c9f7
3
+ size 19201249
checkpoint-124000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2633eef2d71561d41fe95b7d988a3162e5595a11da524fc9fbf1907946ab594
3
+ size 5944