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
ff577a2
·
verified ·
1 Parent(s): 5b3fcd1

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-180000/config.json +79 -0
  3. checkpoint-180000/generation_config.json +6 -0
  4. checkpoint-180000/model.safetensors +3 -0
  5. checkpoint-180000/optimizer.pt +3 -0
  6. checkpoint-180000/rng_state_0.pth +3 -0
  7. checkpoint-180000/rng_state_1.pth +3 -0
  8. checkpoint-180000/rng_state_10.pth +3 -0
  9. checkpoint-180000/rng_state_11.pth +3 -0
  10. checkpoint-180000/rng_state_12.pth +3 -0
  11. checkpoint-180000/rng_state_13.pth +3 -0
  12. checkpoint-180000/rng_state_14.pth +3 -0
  13. checkpoint-180000/rng_state_15.pth +3 -0
  14. checkpoint-180000/rng_state_2.pth +3 -0
  15. checkpoint-180000/rng_state_3.pth +3 -0
  16. checkpoint-180000/rng_state_4.pth +3 -0
  17. checkpoint-180000/rng_state_5.pth +3 -0
  18. checkpoint-180000/rng_state_6.pth +3 -0
  19. checkpoint-180000/rng_state_7.pth +3 -0
  20. checkpoint-180000/rng_state_8.pth +3 -0
  21. checkpoint-180000/rng_state_9.pth +3 -0
  22. checkpoint-180000/scheduler.pt +3 -0
  23. checkpoint-180000/trainer_state.json +3 -0
  24. checkpoint-180000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -83,3 +83,4 @@ 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
84
  checkpoint-178000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
85
  checkpoint-179000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
83
  checkpoint-171000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
84
  checkpoint-178000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
85
  checkpoint-179000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
86
+ checkpoint-180000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-180000/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-180000/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-180000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92ed8410d5f0ff6e1e3ef726e9b6891134d464afb39277a28279ab29580246ea
3
+ size 4538452374
checkpoint-180000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9beb933156503d8323fec25a5ca4ee235009c7d657dcb330b6c533cbf5bc2c55
3
+ size 6630551906
checkpoint-180000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2dc1a0890cd87df78b4d86531e74bc259ee103ffdee32b0f7b984e1e1a278771
3
+ size 15984
checkpoint-180000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3eddc4c42769acbdea816439391ff63da9019ef99728316c5cc2beeaa14954c8
3
+ size 15984
checkpoint-180000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ceaf025a3001bdf68cd2f7e4714a100ae0d76569701ef5eac0319fe3d7a06c17
3
+ size 15997
checkpoint-180000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7737e5dae5b6c0f5d45c3ec93b247f3f017e13ee1c1a28725ea85657d272d42a
3
+ size 15997
checkpoint-180000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62a42a77e960173c1dfd57007fc61dcc579bf37337bc8c4c1b3e055622ad0728
3
+ size 15997
checkpoint-180000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3af879d6fb25cf8ed0890d1f945399f5452c2165482595a391aadedb99ffb35
3
+ size 15997
checkpoint-180000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88bf0d89b1d3a3d89bcc1240387c6f80fc196b8583f096874e7091d71ba5a44b
3
+ size 15997
checkpoint-180000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7f1e33032946b0d4b4f0caf91268fe59b23ac10ed2eaf0ec45f512864ed58b4
3
+ size 15997
checkpoint-180000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e524375346722481718bb111aa79019ff8f52dedce7d1b964cb0514ce3d3cad
3
+ size 15984
checkpoint-180000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc15faa1423728903a20ae2d6a15e50e8bfa87050173ef0d2aadf1edc076b402
3
+ size 15984
checkpoint-180000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52042651205654b00edcad22979982e312ebe79f0810949f41ebd5762f15b8e5
3
+ size 15984
checkpoint-180000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b1328b355b3c6d1487d94e3c302a9372cc656a16289e12dd9c5659e97e47940
3
+ size 15984
checkpoint-180000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e663922e693c0d6a44ce1dd9539e155aef140f3f1e60cae63070af582b0bdcc5
3
+ size 15984
checkpoint-180000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9da341623042ffbb523b772e5f5c1b1901b48873b60bddfe39e9bfaebc33a224
3
+ size 15984
checkpoint-180000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:471765553460540278028e2f13486dc2366d3e89ac4de16bcd3d4f05970a1068
3
+ size 15984
checkpoint-180000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:334f97a02f4ad4e8194e5bcf0d9b5f92bd863d8ff06110e230e8487402f51ca7
3
+ size 15984
checkpoint-180000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28776a9c3de5a32cbf30f46040d1cd3eed38c068e90db3841c12b18e0a5bf5f0
3
+ size 1064
checkpoint-180000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b61579aae1a00dca0250872d5c6b7cd5075e3b50f91de3adb0de6642a5d3edf
3
+ size 27918773
checkpoint-180000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:035b609097a9bd292842ad7ce06a249f64379369aa40aae1bf7cd989f839873a
3
+ size 6008