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
d215360
·
verified ·
1 Parent(s): 411fa9c

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-25000/config.json +79 -0
  2. checkpoint-25000/generation_config.json +6 -0
  3. checkpoint-25000/model.safetensors +3 -0
  4. checkpoint-25000/optimizer.pt +3 -0
  5. checkpoint-25000/rng_state_0.pth +3 -0
  6. checkpoint-25000/rng_state_1.pth +3 -0
  7. checkpoint-25000/rng_state_10.pth +3 -0
  8. checkpoint-25000/rng_state_11.pth +3 -0
  9. checkpoint-25000/rng_state_12.pth +3 -0
  10. checkpoint-25000/rng_state_13.pth +3 -0
  11. checkpoint-25000/rng_state_14.pth +3 -0
  12. checkpoint-25000/rng_state_15.pth +3 -0
  13. checkpoint-25000/rng_state_2.pth +3 -0
  14. checkpoint-25000/rng_state_3.pth +3 -0
  15. checkpoint-25000/rng_state_4.pth +3 -0
  16. checkpoint-25000/rng_state_5.pth +3 -0
  17. checkpoint-25000/rng_state_6.pth +3 -0
  18. checkpoint-25000/rng_state_7.pth +3 -0
  19. checkpoint-25000/rng_state_8.pth +3 -0
  20. checkpoint-25000/rng_state_9.pth +3 -0
  21. checkpoint-25000/scheduler.pt +3 -0
  22. checkpoint-25000/trainer_state.json +0 -0
  23. checkpoint-25000/training_args.bin +3 -0
checkpoint-25000/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-25000/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-25000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b7ebf9db219dce4655868f8d7825330df720e32fbecd23aaa3baabbdff717d5
3
+ size 4559801712
checkpoint-25000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3161e8bd7e4c2d88d4c05d7909ff030c86a2fe2f41ef2fb8cfb588110cfb6eab
3
+ size 6628781666
checkpoint-25000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b172da60145d4d5316f449a5c9a6a8087f7b2f3a4ecff54e3c305c6761a9821f
3
+ size 15984
checkpoint-25000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:298095e0be4c7fd58899614865e1abea04f4a1eeee0b23c21b66df7d33d275c9
3
+ size 15984
checkpoint-25000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5263d84fb60e3f692d76df599b617b63e55aa5d78bc8add17316f2b52b22aa79
3
+ size 15997
checkpoint-25000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f6b03be4cb2a2839a6c2b2d5782171730795149a7f8440776b6c442936a393e
3
+ size 15997
checkpoint-25000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f826b262174170d142c35d43dd0b62ca48c2afcfc2c8a266c90ac886f66d0193
3
+ size 15997
checkpoint-25000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bc386b8963876021f0e50b82c3e4f10dd7612bac2cdbec573c7a9c1f8945439
3
+ size 15997
checkpoint-25000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37ebead65ca0aaa006d7a1546a48daa6dc98816262f98380addb89ed766e70ed
3
+ size 15997
checkpoint-25000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcbfa4a387c12a3a5ff77735cd65d6a5ac2748f67ba1efabd0ff0ea8fa82359d
3
+ size 15997
checkpoint-25000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b33b7666ea6afc2ccfd42f4ea4e9501429fcc038bc12af19d1aa282d93cea50d
3
+ size 15984
checkpoint-25000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b2b7c2b2ac4529190028bc7d05deea4e2d01d5e7562099bbe65451e83f1df9f
3
+ size 15984
checkpoint-25000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:864a65aa58a2e46d25b5d9dd3b6d31fce8207f114b398fa90e064f0deaf0142c
3
+ size 15984
checkpoint-25000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f06975d812d622ccd29494121eee2db1023ec8cab7ccfe27ae4d7597cd747480
3
+ size 15984
checkpoint-25000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d8a48343543a4af56070c0dbcaf23de4abcefb1d0ff6637534956897963f5c2
3
+ size 15984
checkpoint-25000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:286566a0bd1ea8a0282fc1bbdfb521d39a02df9fe70fe6e0d27b0d2a62bfdddd
3
+ size 15984
checkpoint-25000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:010b0fd039939107f72e6269bdcac6d1fd8fae27d45687c67126f5b22532275d
3
+ size 15984
checkpoint-25000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7a1d994e65f7024dda2b83b371314df4e7ac7dd5aa97e4cd29569127f044eae
3
+ size 15984
checkpoint-25000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d4366fb0c40bf7b5d0258969644561831f3926db0473cbee3cdaab374f24646
3
+ size 1064
checkpoint-25000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-25000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1082998a0e1b1eccbf00ed00b26e8e533d335d6c23608114aedab4e2976d2568
3
+ size 5944