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
a98b755
·
verified ·
1 Parent(s): 9832a9f

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-110000/config.json +79 -0
  3. checkpoint-110000/generation_config.json +6 -0
  4. checkpoint-110000/model.safetensors +3 -0
  5. checkpoint-110000/optimizer.pt +3 -0
  6. checkpoint-110000/rng_state_0.pth +3 -0
  7. checkpoint-110000/rng_state_1.pth +3 -0
  8. checkpoint-110000/rng_state_10.pth +3 -0
  9. checkpoint-110000/rng_state_11.pth +3 -0
  10. checkpoint-110000/rng_state_12.pth +3 -0
  11. checkpoint-110000/rng_state_13.pth +3 -0
  12. checkpoint-110000/rng_state_14.pth +3 -0
  13. checkpoint-110000/rng_state_15.pth +3 -0
  14. checkpoint-110000/rng_state_2.pth +3 -0
  15. checkpoint-110000/rng_state_3.pth +3 -0
  16. checkpoint-110000/rng_state_4.pth +3 -0
  17. checkpoint-110000/rng_state_5.pth +3 -0
  18. checkpoint-110000/rng_state_6.pth +3 -0
  19. checkpoint-110000/rng_state_7.pth +3 -0
  20. checkpoint-110000/rng_state_8.pth +3 -0
  21. checkpoint-110000/rng_state_9.pth +3 -0
  22. checkpoint-110000/scheduler.pt +3 -0
  23. checkpoint-110000/trainer_state.json +3 -0
  24. checkpoint-110000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -50,3 +50,4 @@ checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
50
  checkpoint-101000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoint-108000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
52
  checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
50
  checkpoint-101000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
51
  checkpoint-108000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
52
  checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
53
+ checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-110000/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-110000/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-110000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d918ebe9b0a1dea6617b9c908efce8ea416d8777ed6181cde574e8129ed6c78
3
+ size 4538452374
checkpoint-110000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78565ef84aedb68683187738da132ab39f0ee67bd0c810d0ea36299cb30e81ae
3
+ size 6630551906
checkpoint-110000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03f9c823911e3d73101cfda2d0c41068d26046cfabb6095dfd65551c232ea84f
3
+ size 15984
checkpoint-110000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc8450151b8113997ffb99cd2a121a7d1d77c1baa1020cc303ad8bd7ad1b359c
3
+ size 15984
checkpoint-110000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45b5379ae4c3751c2512aeecf39534005f1297c6ad0ced955558bcd3a71e78f2
3
+ size 15997
checkpoint-110000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a5e468928a451cd6e97e8d5730afbc0711a735da59589c0e44f41f4019610f4
3
+ size 15997
checkpoint-110000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11039054f227978cfde030aa7014027267f6899bf74e88276dda0a194ef745e0
3
+ size 15997
checkpoint-110000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41f90ab6b76ec57a7364ddfa4ddb8f2bc8a6c990107984443b7ee3d7432e3757
3
+ size 15997
checkpoint-110000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dc0f93fd51633eac28bee767b52aff421a97aef79b195e7431347794d29d622
3
+ size 15997
checkpoint-110000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1f59445c70c96b0f6f4d09cc13dcdf705e2c5d533ed35b1430b8f9119cd6838
3
+ size 15997
checkpoint-110000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3823f910318efeabe8bc0bd6182ff297f1e3ca3e23db258ea2151be75e2ca424
3
+ size 15984
checkpoint-110000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:536f84daec9df159f893338e0459a12de7ed1f68529eeb76a7deebf2b67eb835
3
+ size 15984
checkpoint-110000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ea3cab0b87ab57808546bffbfdef439551c59c83811d764aacb57da7072dbb4
3
+ size 15984
checkpoint-110000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79c0e86b6dc65fe1c5bdc2458e52e4668e8c8f0ae015eb02f4466ea809be08c8
3
+ size 15984
checkpoint-110000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad206c6fdc245b52e7894d17ffcf825bc803a84c55af5d8cc83a068f4efe333e
3
+ size 15984
checkpoint-110000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07677e4a08f31a6c167fa1eac29d4810d6ff36209b7e91c321c3bc9ba7e86369
3
+ size 15984
checkpoint-110000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e489ff0ff520489a614bb6f372516652f4978d44f415aec6d2317f8f2ef8932
3
+ size 15984
checkpoint-110000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9b42c774ae89e23537295baa706557a336475103cb66f38bc618afce56da384
3
+ size 15984
checkpoint-110000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33764350824c3a31308d8cd939392616f848db881043a4e10d2994e9afc099a4
3
+ size 1064
checkpoint-110000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31437de4fc757d8528f7673123d90383036693ba6f77eb4ddca5db80892958b5
3
+ size 17022034
checkpoint-110000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25f4b7956ecd41dc97ddcc5ee4f9576df83912e2538abf0f4ad824a680639f2f
3
+ size 5944