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
b7a9230
·
verified ·
1 Parent(s): 55ab221

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
@@ -60,3 +60,4 @@ checkpoint-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
60
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-146000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
60
  checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-146000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-148000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
+ 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": 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-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:c677a4460ba854790aacd94cc0779dd02fd16e960b8821ec73d3fc98b37de360
3
+ size 4539487822
checkpoint-150000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:113bdb91335a704c230294446829225fdb725a548079c7f80c4a2fbb7d7751de
3
+ size 6632345482
checkpoint-150000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70bea9a714441ea75128ab273ed6237ce5492a4eb02dffaef0143dfc3e8d748e
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:f24409cdfab9ee5d49f96ceddee5989bbd530f9e7d80aaf1c6efbf417dc13d6e
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:dc46bb288af7a11a3e6626eea4a274f1eacbbc910cf0a5d63d988fe97a6c1088
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 sha256:1f7f06c98613eaee3e141135487fabde2380e120f4ffca78215ccab02292675e
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:ae1c3bdeb1589ca9719b4c05ae57f8875db9ddf7990148f466e1725cb2c529f4
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 sha256:6d300c9d8aa72796853e1a0f21112501b38acdb6fec6c160a4e720633bd05450
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:d5c193db2571e6ab9da4f5540ce5ae71fc226fff088220119dc344a53c5d752a
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:ab0d8b028582cf4e2d41294035656fc99a9916047e916651e40dcfbdaad20957
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 sha256:947f620f36b6663034773eb7d4fd15524148969bc37b5ba2bb29d36c19aada7e
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:c1c3eb8d572894fc18e47c4e2a45d6b2e33d5df6148bff18d511a513256eafa1
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 sha256:2a5159254a72626eaa34b8a969e791d2eca5499040a6e691c1c9793601222d9e
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:da2ddcd52d73c11ca51be5d383e0c73dae058d6acdbf64e5e0ceeaceac0d2faa
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:d5b213015fe91885ee7d0da1c6bad7dcd61d6e42ff630be92b259478fed3d189
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 sha256:4ddd2129f7b039d402267e1adae661a59f97615699108c61ed39bfae02d56e87
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:a910862b2d790862c4b3f233aa68170b64c5c625a5abc52e06af665a2a844d83
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 sha256:32098cb45a455bf3a7fe61791aa373e1964957577909db103b176420c89ae528
3
+ size 15984
checkpoint-150000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb8596898f44287ea3811aa1699eacd7316cc0974a138e75d6f4825669fe5dc1
3
+ size 1064
checkpoint-150000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6accbb5777eaf21e7497a5f59c2ed0b07a8e00cf40b121000731bf057161c4ac
3
+ size 23183979
checkpoint-150000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e61d1d31fca7ac569538321342e23d8e85044bd5092d1e9ede0599b8ee2dd168
3
+ size 6008