Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-ae-f32c32-1.0-768_patch-1_group-7

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
a3a2ee3
·
verified ·
1 Parent(s): 9170744

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-13000/config.json +79 -0
  2. checkpoint-13000/generation_config.json +6 -0
  3. checkpoint-13000/model.safetensors +3 -0
  4. checkpoint-13000/optimizer.pt +3 -0
  5. checkpoint-13000/rng_state_0.pth +3 -0
  6. checkpoint-13000/rng_state_1.pth +3 -0
  7. checkpoint-13000/rng_state_10.pth +3 -0
  8. checkpoint-13000/rng_state_11.pth +3 -0
  9. checkpoint-13000/rng_state_12.pth +3 -0
  10. checkpoint-13000/rng_state_13.pth +3 -0
  11. checkpoint-13000/rng_state_14.pth +3 -0
  12. checkpoint-13000/rng_state_15.pth +3 -0
  13. checkpoint-13000/rng_state_2.pth +3 -0
  14. checkpoint-13000/rng_state_3.pth +3 -0
  15. checkpoint-13000/rng_state_4.pth +3 -0
  16. checkpoint-13000/rng_state_5.pth +3 -0
  17. checkpoint-13000/rng_state_6.pth +3 -0
  18. checkpoint-13000/rng_state_7.pth +3 -0
  19. checkpoint-13000/rng_state_8.pth +3 -0
  20. checkpoint-13000/rng_state_9.pth +3 -0
  21. checkpoint-13000/scheduler.pt +3 -0
  22. checkpoint-13000/trainer_state.json +0 -0
  23. checkpoint-13000/training_args.bin +3 -0
checkpoint-13000/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": 1,
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-in-1.0-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-13000/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-13000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fca54d9e069e71be8a03f658eb2b520466f9247a57d42e71a2bcca999995b975
3
+ size 4559801712
checkpoint-13000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21e89321b8890580e192b1a0c4aadd3d86a05fe19f9633854adfa151b966f405
3
+ size 6628781666
checkpoint-13000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d9ba7b5bbb05d74e8647f998e3dc0794c582335ef7ba498b04abbc73bbe4478
3
+ size 15984
checkpoint-13000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48d355912bf08e87cb7e393bb3554805528392a2be528defe566a13c709e53a6
3
+ size 15984
checkpoint-13000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:134bb2504503a2c02585571a1253c1c844e24759069fa621f2f82290624bf979
3
+ size 15997
checkpoint-13000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd13521956031948068a8a5fc99fbb20876ddb9416d893c60f40ae1b5c8fda8c
3
+ size 15997
checkpoint-13000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef90eb8434a9abf961ced28f09db49aacfee9466271187fccede94203d04d9c8
3
+ size 15997
checkpoint-13000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4ca4aa5e31145fe7d97fc253b1d362c12f5f83a0e9a28425b0124f1e0522e31
3
+ size 15997
checkpoint-13000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33d3a39466b4d87c9a258d193e7d2b23365e20ca4a0d0ffece518a7fc82296d9
3
+ size 15997
checkpoint-13000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51e48bbcb4b0d696c8e28e4d1dcf1f210d5710e382e514c3347932e4c311cdc0
3
+ size 15997
checkpoint-13000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20f1513378d94cc11d3184471d2f0f87265ce4e1005032f6389225f087795442
3
+ size 15984
checkpoint-13000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03e94bcccc0cd491dcab13040a32e802ed2cd7ca176242e00c3918b083503421
3
+ size 15984
checkpoint-13000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1490941203f27576f84f01302a7ca5590ea290352701ca6522a0db9b025d0d91
3
+ size 15984
checkpoint-13000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:520b64a933a9c0f013656c3b9239fa97dbe82a1eb2ff36caef01a72b8e1d99c1
3
+ size 15984
checkpoint-13000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b78e09f3cc429b6fade49f210e994eb5a78711ff81fb15e591cd52529f22f0a
3
+ size 15984
checkpoint-13000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3573785b84d6bb0d80e38ccafc761fb149846bb464dcdee5fccedbe1a84b5778
3
+ size 15984
checkpoint-13000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c67ac7bf8694759926cebc6386c2424ad78dd602c58a35125d768b22a26916ef
3
+ size 15984
checkpoint-13000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ab860770d2b031111d4fd612695b7480a4c820157537b1b1f9f50ee1e615f54
3
+ size 15984
checkpoint-13000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9a5dbd38a5bcb4cc58eda118d253102ac321e401868e576d45013398dad778f
3
+ size 1064
checkpoint-13000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-13000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d651f5bb99f85a60d02309d0be08d2460290b4709b30bf924509259f2f349375
3
+ size 5944