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
5d77ec8
·
verified ·
1 Parent(s): 1e916db

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-171000/config.json +79 -0
  3. checkpoint-171000/generation_config.json +6 -0
  4. checkpoint-171000/model.safetensors +3 -0
  5. checkpoint-171000/optimizer.pt +3 -0
  6. checkpoint-171000/rng_state_0.pth +3 -0
  7. checkpoint-171000/rng_state_1.pth +3 -0
  8. checkpoint-171000/rng_state_10.pth +3 -0
  9. checkpoint-171000/rng_state_11.pth +3 -0
  10. checkpoint-171000/rng_state_12.pth +3 -0
  11. checkpoint-171000/rng_state_13.pth +3 -0
  12. checkpoint-171000/rng_state_14.pth +3 -0
  13. checkpoint-171000/rng_state_15.pth +3 -0
  14. checkpoint-171000/rng_state_2.pth +3 -0
  15. checkpoint-171000/rng_state_3.pth +3 -0
  16. checkpoint-171000/rng_state_4.pth +3 -0
  17. checkpoint-171000/rng_state_5.pth +3 -0
  18. checkpoint-171000/rng_state_6.pth +3 -0
  19. checkpoint-171000/rng_state_7.pth +3 -0
  20. checkpoint-171000/rng_state_8.pth +3 -0
  21. checkpoint-171000/rng_state_9.pth +3 -0
  22. checkpoint-171000/scheduler.pt +3 -0
  23. checkpoint-171000/trainer_state.json +3 -0
  24. checkpoint-171000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -80,3 +80,4 @@ 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
 
 
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
83
+ checkpoint-171000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-171000/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-171000/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-171000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88b204c65eb505622618ac1aad4ce09aee5f1c16a41efd81be56ba7a4911d71b
3
+ size 4538452374
checkpoint-171000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:036abf7759d0243c52af51ebb8ad9869cf41c6e5869141592e69d182b287db38
3
+ size 6630551906
checkpoint-171000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9946c17a03d24b8ea8920f662da7f7ebdcff1d42fa2becdecd7f4efa59a586c
3
+ size 15984
checkpoint-171000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f62bc020e1f61797b5752bcbaed8c4ab697e8cbed3df03c315564acae3f766d8
3
+ size 15984
checkpoint-171000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9aea43d306237c3be3f0fe616ce10ac40179e12c3eb2dd2861a27b0623281217
3
+ size 15997
checkpoint-171000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52c40d16c7a9ddab905ea0b598cad6d5c9aa190f1663675cd72931eac251a9f6
3
+ size 15997
checkpoint-171000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93d49bd686097bc4b25ec8d2f3bd7d60a28981c379c0ad806c611a5b19935e9d
3
+ size 15997
checkpoint-171000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7adbe5aaffe21e324617d2b63e4192d9da92dceb95a63bf51c129ce213b73f85
3
+ size 15997
checkpoint-171000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c97088d21ec4a3d03b300268a70d02415d5888ba6e8d14b80487d97b93fb8a5
3
+ size 15997
checkpoint-171000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d14d81f230ff34593a5d9fb9ed747aa343a96fc5a23a7164931c84235f7d8275
3
+ size 15997
checkpoint-171000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f976fd29ceca77c9d6eb00bc25973c8c9bec7f1db36a3116c3180da0d7fe231
3
+ size 15984
checkpoint-171000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b681aecaeb8b56c676ddb62754014a321be3475de0e723b56bab131a213ebc6
3
+ size 15984
checkpoint-171000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e454486b95bd189b94318a0d69b5a75c7e4f79bca8b5d8bea9af4cf8815a99ec
3
+ size 15984
checkpoint-171000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:441e92f5bd5ba7415893643ff81934c956a72b395457140a48f7ada82b11ef3c
3
+ size 15984
checkpoint-171000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:925a3178631c6fa8a9382ce2f582f8b1dd3f2c935175d15c47076b0e13caf026
3
+ size 15984
checkpoint-171000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cdbad98bb9e0ea29486887d79b7fe7bda5a9964ff7a0db1088fa1c2f2ae52cff
3
+ size 15984
checkpoint-171000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ebb55c793177b2280cf894f15b6b921d6f1be05f82374a9bb25843c6a6249f7
3
+ size 15984
checkpoint-171000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50fb8d65ad07056b4cc504b23d8f921111a346e92ddd3bafbbf6b7d5458cf756
3
+ size 15984
checkpoint-171000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa895eb03266d395010df45fbcb9d6e732bfd4bd36b5f961025777e6c7c95146
3
+ size 1064
checkpoint-171000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63a0055f8934f21954167ba683a91ed7de1f199282d0aa65292e1a8ae0ef315c
3
+ size 26520342
checkpoint-171000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db32f51058b4dfd37dd19a4e4ea10e2c2235e37dafb7e35fef8b91089225dedc
3
+ size 5944