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
f0719b3
·
verified ·
1 Parent(s): 7f38603

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-150000/config.json +79 -0
  3. checkpoint-150000/generation_config.json +6 -0
  4. checkpoint-150000/model.safetensors +3 -0
  5. checkpoint-150000/optimizer.pt +3 -0
  6. checkpoint-150000/rng_state_0.pth +3 -0
  7. checkpoint-150000/rng_state_1.pth +3 -0
  8. checkpoint-150000/rng_state_10.pth +3 -0
  9. checkpoint-150000/rng_state_11.pth +3 -0
  10. checkpoint-150000/rng_state_12.pth +3 -0
  11. checkpoint-150000/rng_state_13.pth +3 -0
  12. checkpoint-150000/rng_state_14.pth +3 -0
  13. checkpoint-150000/rng_state_15.pth +3 -0
  14. checkpoint-150000/rng_state_2.pth +3 -0
  15. checkpoint-150000/rng_state_3.pth +3 -0
  16. checkpoint-150000/rng_state_4.pth +3 -0
  17. checkpoint-150000/rng_state_5.pth +3 -0
  18. checkpoint-150000/rng_state_6.pth +3 -0
  19. checkpoint-150000/rng_state_7.pth +3 -0
  20. checkpoint-150000/rng_state_8.pth +3 -0
  21. checkpoint-150000/rng_state_9.pth +3 -0
  22. checkpoint-150000/scheduler.pt +3 -0
  23. checkpoint-150000/trainer_state.json +3 -0
  24. checkpoint-150000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -71,3 +71,4 @@ checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
71
  checkpoint-141000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
72
  checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
73
  checkpoint-149000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
71
  checkpoint-141000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
72
  checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
73
  checkpoint-149000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
74
+ checkpoint-150000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-150000/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-150000/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-150000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7ee55fbf09c3d78697e79a9fa605bab80856320205b3319c07770e9769f8839
3
+ size 4538452374
checkpoint-150000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40e3ccc8de7370089e3e9b35b1ea9967ae593ce5ebd10ba03a2ee12ee5c622e7
3
+ size 6630551906
checkpoint-150000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931833d505e0efc32dfe7a99186e8ae3b62a33726b0b2b09f4cc0698040d571e
3
+ size 15984
checkpoint-150000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cda42d7a43d585de85d081969f06beb2a3064714b2701690c68258e974dd806
3
+ size 15984
checkpoint-150000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb0296fdd9d198bfb23f8ccefd2f452a59dda17ccca3f216748c43cdf02ab4b9
3
+ size 15997
checkpoint-150000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b341f8012378604c288d8b3ac83f8cfb4fe89f3dec8e785b875ca9ea4856b63
3
+ size 15997
checkpoint-150000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f52f2fbda7134340d5631f6d949c12efae5f5c73e44613d68ab9a4d60878121e
3
+ size 15997
checkpoint-150000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f9d17cb3e98691f338510ffff0da8fd4dce8d4828c5918f57715f8d80d56dfe
3
+ size 15997
checkpoint-150000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fee869f3e100978287f08f3cad777094f39be59cf7cc3ee5405ba83eb883907
3
+ size 15997
checkpoint-150000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ecbb8fa995e5492fee7953631aef53b69ec974cdfc65e0be6277a71e1f25ee1
3
+ size 15997
checkpoint-150000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0f3bef59b762dde58c0bf0fae983d8e60761f410bed985918900467aa1c5863
3
+ size 15984
checkpoint-150000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29d30758b32842e190e2971cc6b3a63be457fd3a1f0da438f14ff6bfc2278392
3
+ size 15984
checkpoint-150000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91b20fb80f2a2c23da3a8ffaee3ed4ffca97276a9753d2cacffbaf43b9c7ddb4
3
+ size 15984
checkpoint-150000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32af79c0da569abd75f1c49fc16f26a3e6cd2558fb26e45d42c2a01da083cff4
3
+ size 15984
checkpoint-150000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07dfe5b16d8e4bd91496185d08b692e96d6a5274eb5cb4858fd0f4e7f311b64d
3
+ size 15984
checkpoint-150000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae830c0cb1e9105bfb74f021074db7f1c62696d67c2c7b30c1103dbbb8f8f549
3
+ size 15984
checkpoint-150000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd28bd01f2b331c1d936669b3a70eb17281c856a1c286c20ba5617044346bee1
3
+ size 15984
checkpoint-150000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93728083ac3abb0266a72852cf9f4e7a13f8e321b48b091c2c2a70fc869e5a74
3
+ size 15984
checkpoint-150000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b85e110a22ad5aff4c0cb08ad753097dafd07aaff142c7ecfccffcb2d513627c
3
+ size 1064
checkpoint-150000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83dadd56dc3c1080e8e070580805fb243f3456900e5ed4e0b41b2eb9972a183e
3
+ size 23251690
checkpoint-150000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57eeba44bb319121fea8ce8ee4753bd775b1bf8770da408b2c2b5d549905064f
3
+ size 6008