Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_aug_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
d2f1929
·
verified ·
1 Parent(s): edf9f00

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-162000/config.json +79 -0
  3. checkpoint-162000/generation_config.json +6 -0
  4. checkpoint-162000/model.safetensors +3 -0
  5. checkpoint-162000/optimizer.pt +3 -0
  6. checkpoint-162000/rng_state_0.pth +3 -0
  7. checkpoint-162000/rng_state_1.pth +3 -0
  8. checkpoint-162000/rng_state_10.pth +3 -0
  9. checkpoint-162000/rng_state_11.pth +3 -0
  10. checkpoint-162000/rng_state_12.pth +3 -0
  11. checkpoint-162000/rng_state_13.pth +3 -0
  12. checkpoint-162000/rng_state_14.pth +3 -0
  13. checkpoint-162000/rng_state_15.pth +3 -0
  14. checkpoint-162000/rng_state_2.pth +3 -0
  15. checkpoint-162000/rng_state_3.pth +3 -0
  16. checkpoint-162000/rng_state_4.pth +3 -0
  17. checkpoint-162000/rng_state_5.pth +3 -0
  18. checkpoint-162000/rng_state_6.pth +3 -0
  19. checkpoint-162000/rng_state_7.pth +3 -0
  20. checkpoint-162000/rng_state_8.pth +3 -0
  21. checkpoint-162000/rng_state_9.pth +3 -0
  22. checkpoint-162000/scheduler.pt +3 -0
  23. checkpoint-162000/trainer_state.json +3 -0
  24. checkpoint-162000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -64,3 +64,4 @@ checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
  checkpoint-156000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  checkpoint-158000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
66
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
64
  checkpoint-156000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
65
  checkpoint-158000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
66
  checkpoint-160000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
67
+ checkpoint-162000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-162000/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": true,
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-162000/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-162000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46332213bac843ff87d5da92bf6b7d0b68c43b01f7e7e4b4c8407eb9671816fc
3
+ size 4539487822
checkpoint-162000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a74dc6bb805b31a95aa68d0eb2c0aed271254621bb37c2111e12d455b30fb7e
3
+ size 6632345482
checkpoint-162000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8551c41d714bbefb805ac526419ebf95db867dd5b83b4264924fa4ef36e20b3a
3
+ size 15984
checkpoint-162000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a3fd21209f32a378198386f46d56d1db5622a4edc2ada9032876aedd80fb75d
3
+ size 15984
checkpoint-162000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f91e4892e65adf49b07f6718d71913b27431cd99c3295eb34eb6aaafc828c470
3
+ size 15997
checkpoint-162000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c9d3039853c053cc0b604662eff22284d12e838a8ff90a696371eb3d7eff235
3
+ size 15997
checkpoint-162000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ec1621f36500dab6c4f69ee16d17d868e938101fd3540f2cbc600dd9f661c3a
3
+ size 15997
checkpoint-162000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac6a755c7488c064e29396153f66a78da6ebdd0514543f632a3afe9214a02449
3
+ size 15997
checkpoint-162000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a103e95b4251be539bf9c4f27c0888387c86459e773e59e673cc0e5854a1db74
3
+ size 15997
checkpoint-162000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1fb90a2f288d0b63b1e6f91c5f6d74b98c7b05d54cfcb72528f7656580cd94c
3
+ size 15997
checkpoint-162000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2f9afe3170b273ed2dbc303779c81aeb5010828361fe9cd585a497cc13257d3
3
+ size 15984
checkpoint-162000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3e24a3db62e3469ba5b6182d45dd9bab49a34f9ec3c7b4b9257c2b25dc45245
3
+ size 15984
checkpoint-162000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dc0d0e5480f7c5d6ec5588dc4174a19f5d1abd8e972fbd90785ead9c8a60337
3
+ size 15984
checkpoint-162000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7803ecd7cf630116247c747d76c6a991bf7c6075155cc5e72b5c06b6158a638
3
+ size 15984
checkpoint-162000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b1cdffc50858d97c9a83a51917a7bfef19ef6fe173601c088ae98f23f27829c
3
+ size 15984
checkpoint-162000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60fbc6f13fa5f867f07648effad26a8d9403dfadbe6a5865b2692c821a84bc0e
3
+ size 15984
checkpoint-162000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f160c9d068b69cc756dd23a041cd5d3b11bf493f18b40e67d151dec8b5273633
3
+ size 15984
checkpoint-162000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efcfcdb99f8694e0c96fdba2cc3b711887ed2668aecfd8f462e0e30dc55de615
3
+ size 15984
checkpoint-162000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcf3f0d15d2976539fd8779a4d6ea51993435b5f907e1390e560860cee04517b
3
+ size 1064
checkpoint-162000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8541742c760a2bf0dfe1646dbc61a3db5a8ee03a2262b7bbfc633d20229817f
3
+ size 25046101
checkpoint-162000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfbceb40fbd2389ef65045bc07d28fab008436a329649f9f023e41014c905036
3
+ size 6008