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
1e916db
·
verified ·
1 Parent(s): b417a3e

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-170000/config.json +79 -0
  3. checkpoint-170000/generation_config.json +6 -0
  4. checkpoint-170000/model.safetensors +3 -0
  5. checkpoint-170000/optimizer.pt +3 -0
  6. checkpoint-170000/rng_state_0.pth +3 -0
  7. checkpoint-170000/rng_state_1.pth +3 -0
  8. checkpoint-170000/rng_state_10.pth +3 -0
  9. checkpoint-170000/rng_state_11.pth +3 -0
  10. checkpoint-170000/rng_state_12.pth +3 -0
  11. checkpoint-170000/rng_state_13.pth +3 -0
  12. checkpoint-170000/rng_state_14.pth +3 -0
  13. checkpoint-170000/rng_state_15.pth +3 -0
  14. checkpoint-170000/rng_state_2.pth +3 -0
  15. checkpoint-170000/rng_state_3.pth +3 -0
  16. checkpoint-170000/rng_state_4.pth +3 -0
  17. checkpoint-170000/rng_state_5.pth +3 -0
  18. checkpoint-170000/rng_state_6.pth +3 -0
  19. checkpoint-170000/rng_state_7.pth +3 -0
  20. checkpoint-170000/rng_state_8.pth +3 -0
  21. checkpoint-170000/rng_state_9.pth +3 -0
  22. checkpoint-170000/scheduler.pt +3 -0
  23. checkpoint-170000/trainer_state.json +3 -0
  24. checkpoint-170000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -79,3 +79,4 @@ checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
  checkpoint-161000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-168000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
79
  checkpoint-161000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
  checkpoint-168000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
81
  checkpoint-169000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
82
+ checkpoint-170000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-170000/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-170000/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-170000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e93867f32d6e2fb47b3a710f2344dc5157e41a3a384da2b69908f70b0143ae3
3
+ size 4538452374
checkpoint-170000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a6cce29010f36afa140d780342b6daee38f23fb7cc61bf5ef5c6982fc843586
3
+ size 6630551906
checkpoint-170000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5a23bcf47df003c9c46a3a0a7c929ed82987a7dc494c6223062293e669dcd41
3
+ size 15984
checkpoint-170000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a553ffb1fdc704610637c05b84ba2f9706fe5979c49a2657e46aaa1330254ff
3
+ size 15984
checkpoint-170000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:021f7a1d53600f701b09f0c3014adec5fc9a0a79d80d3dada90dcd45ce674eaa
3
+ size 15997
checkpoint-170000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd5282a86fec4151daf47a47ed01cf6ca94b14529f27dd96bc646e7a7d8ba301
3
+ size 15997
checkpoint-170000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:284b1dd2e9bba2a17b0ffb2b7bb77ab499cad48e7c869afce1b17703de101e04
3
+ size 15997
checkpoint-170000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6876f33471c2edd95be716bb3af679edd92d3c8501a58a2cab673989db7d0a7f
3
+ size 15997
checkpoint-170000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c844569a03b154cd132843de041574e219a52d3ff2aee6f21996d883cf04fa5f
3
+ size 15997
checkpoint-170000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:800cc5ad9b10b75cd1b44734a554593ccf6fda13296344d8e6c9852c3a3e4cd6
3
+ size 15997
checkpoint-170000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4596ebaf22d1930a95c0ba36fee9a1adbfd38f4eb635091c2aec4b4c592201bc
3
+ size 15984
checkpoint-170000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecf55c439ddb939ce0579331060f3d7ba5c5622a2e08e40832d4d8dc8430ecac
3
+ size 15984
checkpoint-170000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5689bec4ab8948ac78df5e8016d97ef93c5a9fb7214c6b0392b7a8bdf4abc6b
3
+ size 15984
checkpoint-170000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0553f9fc62081fcfca0ad5e5cf1dc131f66b62a2f664e229cbc27a0c07bb6c00
3
+ size 15984
checkpoint-170000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02fed6dc393c4de0cf978d88f65260a7f552d9991a4eb9b798d2036070181f08
3
+ size 15984
checkpoint-170000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f8f3598b719f964aeb460069f5af87bcb90dec7ee34773e250ff1ac394659fd
3
+ size 15984
checkpoint-170000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e28392666528428ebd26fa1972e9ff5b2c99b5b2a91ce18958ab01b6a14f52d3
3
+ size 15984
checkpoint-170000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a36bbb57b470e289c2923e4e7bf8007321291816733de1701f3411fc53088380
3
+ size 15984
checkpoint-170000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53418078886f31c0993feeddb7d41f245589575ed7c2ae4e89cfde013fe46589
3
+ size 1064
checkpoint-170000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e26def1e9d7d72cc926d5d5a02386749806c84d4ef45c73c46ae2a4c477da67
3
+ size 26364936
checkpoint-170000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0c72be1f87ae766f1c632ef628f5a17769a136305fc882ce6f8f478b8d0348a
3
+ size 5944