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
f3a3ced
·
verified ·
1 Parent(s): ee9e49e

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-114000/config.json +79 -0
  3. checkpoint-114000/generation_config.json +6 -0
  4. checkpoint-114000/model.safetensors +3 -0
  5. checkpoint-114000/optimizer.pt +3 -0
  6. checkpoint-114000/rng_state_0.pth +3 -0
  7. checkpoint-114000/rng_state_1.pth +3 -0
  8. checkpoint-114000/rng_state_10.pth +3 -0
  9. checkpoint-114000/rng_state_11.pth +3 -0
  10. checkpoint-114000/rng_state_12.pth +3 -0
  11. checkpoint-114000/rng_state_13.pth +3 -0
  12. checkpoint-114000/rng_state_14.pth +3 -0
  13. checkpoint-114000/rng_state_15.pth +3 -0
  14. checkpoint-114000/rng_state_2.pth +3 -0
  15. checkpoint-114000/rng_state_3.pth +3 -0
  16. checkpoint-114000/rng_state_4.pth +3 -0
  17. checkpoint-114000/rng_state_5.pth +3 -0
  18. checkpoint-114000/rng_state_6.pth +3 -0
  19. checkpoint-114000/rng_state_7.pth +3 -0
  20. checkpoint-114000/rng_state_8.pth +3 -0
  21. checkpoint-114000/rng_state_9.pth +3 -0
  22. checkpoint-114000/scheduler.pt +3 -0
  23. checkpoint-114000/trainer_state.json +3 -0
  24. checkpoint-114000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -54,3 +54,4 @@ checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
54
  checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
  checkpoint-113000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
54
  checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
  checkpoint-113000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
+ checkpoint-114000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-114000/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-114000/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-114000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a3b69fa319010c6f4282e6f92d05a6bbb750e236aa74eb01a7cfdf6e76eacc8
3
+ size 4538452374
checkpoint-114000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46da24b577c5822b76ddbf8d18c68d954ec1342cc32bcda871faa53f379f63c6
3
+ size 6630551906
checkpoint-114000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38e65418510b24aae92a371d6b50eae28509af6ae4c7f85bd53d4c2f528be460
3
+ size 15984
checkpoint-114000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c32763852cbe39c2e19b3b0be7d388de4f5bafd3627b6510f0a2cd9e9679211
3
+ size 15984
checkpoint-114000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52312001255d452b353698f25c0e79937d067cb87547c48039329e3f549e54f2
3
+ size 15997
checkpoint-114000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad4a37cb8621a808a4ff8e7e630842387b30febc149bb2f8ceff0ae80459bcd6
3
+ size 15997
checkpoint-114000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98b64aa67a9bf0b920ebc6e7835323b2ddca01ea4bc4c33d17d34760045d10d3
3
+ size 15997
checkpoint-114000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48d887d1d423731b1f46b88cd26c73164ea211564eeb91da7ae4e724fb7276fe
3
+ size 15997
checkpoint-114000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3778c70ee598228870d9c2e342ec75f4015c5a805a3f70ac52e0151343490322
3
+ size 15997
checkpoint-114000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:856bfef43bff343582caedec13c961f002f1c15c154c96dc4fcf7cad4b9e0f5e
3
+ size 15997
checkpoint-114000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b30d506f1b001fd29bd031670e6e6cddb711b52dc92b25746d6dabc4210aff33
3
+ size 15984
checkpoint-114000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90166ebcdc900ae9c106714df81334a4344169a6bb5bb30c9fc024de0f187829
3
+ size 15984
checkpoint-114000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13f4a37eec56df3792e81c1d92af8a31939b2aa6a4230b56d1a52019b03d01df
3
+ size 15984
checkpoint-114000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bb2c528bc6932ee9db716b141948be0099617d18054aac66f1f43019b81d3e2
3
+ size 15984
checkpoint-114000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16e3b6a7f64cbbe6723739f742160ccb9f03552432bf94006fe746cedb4cce0a
3
+ size 15984
checkpoint-114000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02fd3faa7610273e34735e8b40bf71ce04cff7329ef85107994367e5eba820aa
3
+ size 15984
checkpoint-114000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e10499cc47d7cf5ef5eeec8281a07646779259c2a3a470027a474644b6e8b339
3
+ size 15984
checkpoint-114000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:184486f1e7502bacca60a670b628a56f7bbdb69168634d1367efba2b129317bc
3
+ size 15984
checkpoint-114000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07b3de2c3dc254eb8d4409466153f7ad0e134c7d6646cc99346b82345675fa6e
3
+ size 1064
checkpoint-114000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d6cedca4b3a3e63d9552a508b024845675266ec209405cbd289b83406bf1f27
3
+ size 17644790
checkpoint-114000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43e7ba1222743d25e382a8c675e15f3f4f75df2138a6ddce9946bebe22b32edd
3
+ size 5944