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
411fa9c
·
verified ·
1 Parent(s): d3a2ec4

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-24000/config.json +79 -0
  2. checkpoint-24000/generation_config.json +6 -0
  3. checkpoint-24000/model.safetensors +3 -0
  4. checkpoint-24000/optimizer.pt +3 -0
  5. checkpoint-24000/rng_state_0.pth +3 -0
  6. checkpoint-24000/rng_state_1.pth +3 -0
  7. checkpoint-24000/rng_state_10.pth +3 -0
  8. checkpoint-24000/rng_state_11.pth +3 -0
  9. checkpoint-24000/rng_state_12.pth +3 -0
  10. checkpoint-24000/rng_state_13.pth +3 -0
  11. checkpoint-24000/rng_state_14.pth +3 -0
  12. checkpoint-24000/rng_state_15.pth +3 -0
  13. checkpoint-24000/rng_state_2.pth +3 -0
  14. checkpoint-24000/rng_state_3.pth +3 -0
  15. checkpoint-24000/rng_state_4.pth +3 -0
  16. checkpoint-24000/rng_state_5.pth +3 -0
  17. checkpoint-24000/rng_state_6.pth +3 -0
  18. checkpoint-24000/rng_state_7.pth +3 -0
  19. checkpoint-24000/rng_state_8.pth +3 -0
  20. checkpoint-24000/rng_state_9.pth +3 -0
  21. checkpoint-24000/scheduler.pt +3 -0
  22. checkpoint-24000/trainer_state.json +0 -0
  23. checkpoint-24000/training_args.bin +3 -0
checkpoint-24000/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-24000/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-24000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff59b654e97c6cfaa2e7632b87ce92368895b5db75881bbbc74447c97c5b77bd
3
+ size 4559801712
checkpoint-24000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8dba675b79ab370012027b3c6e9b8b2892c506b380229174811570299c7a993
3
+ size 6628781666
checkpoint-24000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf26691264fb04a20c08b66888a719be5e4c7c3fa8814dfb11f319be70c6dccf
3
+ size 15984
checkpoint-24000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d3c48a83e18dc3f6ac74fdb82e7c6e88f14f42cb3d0a25bc2a64f2912ab10c1
3
+ size 15984
checkpoint-24000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07b7b5f1442c5023d8deacc513e9b286e32e68ec2b41a07a3af9843c05b17b05
3
+ size 15997
checkpoint-24000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c1dc7cde64332f7849507f49340e982668b5038931c9cbf33067a67d597423a
3
+ size 15997
checkpoint-24000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f81eaa3f9586853e8b11c0a1a42856461ab63f3204d858e5eb129086d10be09c
3
+ size 15997
checkpoint-24000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f134710103c72841c9982006f432a876592fadec4f1b4577429bb0b48e710cc0
3
+ size 15997
checkpoint-24000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4363b513566537b85efa68f19e8adfbaac244dae115c9ebc564a0ea9c30e076
3
+ size 15997
checkpoint-24000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:483e409646e65f67524b6e764cf25d6d569fa430bed25b40357a26a07e5c3f2f
3
+ size 15997
checkpoint-24000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5677953263d7493088804de4b1e2886e3b87dbf9f585cb7e988324b1a7f6f38a
3
+ size 15984
checkpoint-24000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c3e8fa1d4e4b49ce53ea32aaa7e6a216a21f9ab01a9a53bc612a042bca945da
3
+ size 15984
checkpoint-24000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d5c78dcf3bb3bb52078cfa49175c863fbbba5e30c3e3ae2bb813dd7ad28c4d3
3
+ size 15984
checkpoint-24000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afe008f59e1bdfaced24ee9e71697dd6b2f7f12103019381b6def44aa0e48c1d
3
+ size 15984
checkpoint-24000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5df0b7462ebc51a8243bb9a05a1be05eb6f002f31fd8db1144e1383488ec34c
3
+ size 15984
checkpoint-24000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7626da1762b1dd87aea5af5d8393c95d28c487a0ec924cc506fd179fc9a6d18
3
+ size 15984
checkpoint-24000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8286ee4a134be432fcbcd2e9347fc70f6c9d7804546779e1e95f9838558297cf
3
+ size 15984
checkpoint-24000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46409d9b32ea0c11a61d0daba3c91200b5b0511924e8e96645746d8256c06f29
3
+ size 15984
checkpoint-24000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c4ca8d1fc1bb328cef47d23501b6cc59e7895bfc83352e90443703bf935ef80
3
+ size 1064
checkpoint-24000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-24000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88bcfaa9ae584f5771ef2fb35262ef9571236516e267e60ec0ed2f5a659a0f3d
3
+ size 5944