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
08a00e0
·
verified ·
1 Parent(s): 4388d3f

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-226000/config.json +79 -0
  3. checkpoint-226000/generation_config.json +6 -0
  4. checkpoint-226000/model.safetensors +3 -0
  5. checkpoint-226000/optimizer.pt +3 -0
  6. checkpoint-226000/rng_state_0.pth +3 -0
  7. checkpoint-226000/rng_state_1.pth +3 -0
  8. checkpoint-226000/rng_state_10.pth +3 -0
  9. checkpoint-226000/rng_state_11.pth +3 -0
  10. checkpoint-226000/rng_state_12.pth +3 -0
  11. checkpoint-226000/rng_state_13.pth +3 -0
  12. checkpoint-226000/rng_state_14.pth +3 -0
  13. checkpoint-226000/rng_state_15.pth +3 -0
  14. checkpoint-226000/rng_state_2.pth +3 -0
  15. checkpoint-226000/rng_state_3.pth +3 -0
  16. checkpoint-226000/rng_state_4.pth +3 -0
  17. checkpoint-226000/rng_state_5.pth +3 -0
  18. checkpoint-226000/rng_state_6.pth +3 -0
  19. checkpoint-226000/rng_state_7.pth +3 -0
  20. checkpoint-226000/rng_state_8.pth +3 -0
  21. checkpoint-226000/rng_state_9.pth +3 -0
  22. checkpoint-226000/scheduler.pt +3 -0
  23. checkpoint-226000/trainer_state.json +3 -0
  24. checkpoint-226000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -87,3 +87,4 @@ checkpoint-218000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
87
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
88
  checkpoint-222000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
89
  checkpoint-224000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
87
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
88
  checkpoint-222000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
89
  checkpoint-224000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
90
+ checkpoint-226000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-226000/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-226000/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-226000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50dae7ecc7dbfb58a47db0a62192348f5fc343935d1c2575ad2dbcf8490d7251
3
+ size 4539487822
checkpoint-226000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e69011b49539d75eeeebe5aa8b0d5571161f52dbb65371f4fb1ea8f1b9ef3a43
3
+ size 6632345482
checkpoint-226000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11ba91d7a8524749321b2a37afb7a476f43ed725f910795f8de9fe72dc1b30be
3
+ size 15984
checkpoint-226000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cd49c4e4f2943d33ee05f121c55068d200227b71d3f739e1927951f5b53b8ec
3
+ size 15984
checkpoint-226000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27866ee551026cc6863eef4e14517b902903ca97182a92f870b2b469e329d8a5
3
+ size 15997
checkpoint-226000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c427ebbf45c130e3223df01f625ec09abcd0a3e73c91c88e375637024f98cc5
3
+ size 15997
checkpoint-226000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc7a91bb5da65524c62fe1a82b25ad6e1c3957d93f7d4a37d223a70b9c440243
3
+ size 15997
checkpoint-226000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:211c0d97e931ca5cf2bafbe94a04ba77c5449860f96ae8abad7c14f2e62d3561
3
+ size 15997
checkpoint-226000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17f8cc82773f71023bee9c44047dfdc61f438cbe5b4f954ebb025f75892bb8fd
3
+ size 15997
checkpoint-226000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b48c0a02bb9c006413735cadf191ad816fdf108e5610aac755adf4e96b134e39
3
+ size 15997
checkpoint-226000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e21f52486e7037c6a6e2c3f8347b8036bd57088b558f149997edf8d50215d6c
3
+ size 15984
checkpoint-226000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33c37c44e643f79fb49df921897268ac4eb61996838407952b2e9c903c1c33e5
3
+ size 15984
checkpoint-226000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef4ab903fcdce0b08c0e9c581a3070f7e88bfb2a4485f1cc3e24f8a1f5281bcd
3
+ size 15984
checkpoint-226000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f086f13abe6dc38aa74accf43b96abc39aa6bca6ca3f73650a73fb9f206bfe37
3
+ size 15984
checkpoint-226000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad1cb4f5aa6a3812e1b3742b3271ee46896d59da34b4cf32d5e8d9fc26c38a4b
3
+ size 15984
checkpoint-226000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29a74385ba6d212185ec22179c04b3eab1e1939a9a734f92648443f069ecd359
3
+ size 15984
checkpoint-226000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2999b701b0fb4df24620159ab79498da5539c1de4f40934098f4731c3bb294ae
3
+ size 15984
checkpoint-226000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0ae7d03aad5c42cf2ea423ece29143d2c4ac6494bd0255a0aa9c7d72c0b7341
3
+ size 15984
checkpoint-226000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:225ef5a34bd0f54ac17118a158131bfaa004556e68bbd639eb1a3a47df3d1e82
3
+ size 1064
checkpoint-226000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ce0b82e17812f44a135a2703985776e33b1b2c7025fdbba1023cd203eaf4742
3
+ size 34944923
checkpoint-226000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d04a2203f94577d173d7059ef552feaa243fecd0ed76a52a5dd48b35e3b53c7
3
+ size 6008