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
d91660d
·
verified ·
1 Parent(s): f3a3ced

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-115000/config.json +79 -0
  3. checkpoint-115000/generation_config.json +6 -0
  4. checkpoint-115000/model.safetensors +3 -0
  5. checkpoint-115000/optimizer.pt +3 -0
  6. checkpoint-115000/rng_state_0.pth +3 -0
  7. checkpoint-115000/rng_state_1.pth +3 -0
  8. checkpoint-115000/rng_state_10.pth +3 -0
  9. checkpoint-115000/rng_state_11.pth +3 -0
  10. checkpoint-115000/rng_state_12.pth +3 -0
  11. checkpoint-115000/rng_state_13.pth +3 -0
  12. checkpoint-115000/rng_state_14.pth +3 -0
  13. checkpoint-115000/rng_state_15.pth +3 -0
  14. checkpoint-115000/rng_state_2.pth +3 -0
  15. checkpoint-115000/rng_state_3.pth +3 -0
  16. checkpoint-115000/rng_state_4.pth +3 -0
  17. checkpoint-115000/rng_state_5.pth +3 -0
  18. checkpoint-115000/rng_state_6.pth +3 -0
  19. checkpoint-115000/rng_state_7.pth +3 -0
  20. checkpoint-115000/rng_state_8.pth +3 -0
  21. checkpoint-115000/rng_state_9.pth +3 -0
  22. checkpoint-115000/scheduler.pt +3 -0
  23. checkpoint-115000/trainer_state.json +3 -0
  24. checkpoint-115000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -55,3 +55,4 @@ checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
  checkpoint-113000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-114000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
55
  checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
56
  checkpoint-113000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
57
  checkpoint-114000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
58
+ checkpoint-115000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-115000/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-115000/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-115000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc344363d0098da479be0bb198778835423e5ccff2c93337ee61d415f4f2d821
3
+ size 4538452374
checkpoint-115000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bb9238b2719e5c4dc63bcb7cb4205ab72b345505f38e553c4c04218a5032861
3
+ size 6630551906
checkpoint-115000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d11f89ef66ac3d11be93b51f72eb87bdb366e703e7892c9a18611b74178bcf4
3
+ size 15984
checkpoint-115000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5054a8b194ae703b82d563514ac94ffd72ecc17f4572efe7ee8c43202f922442
3
+ size 15984
checkpoint-115000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71b7180b3295df2bc7698703eb739fd91af856e6c39fb6ea9de7d70766681725
3
+ size 15997
checkpoint-115000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:560009cf74c6f022cecfecce903bbc766e5e1caf64778da3661c33908df60344
3
+ size 15997
checkpoint-115000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8d3189a28f7fbddf3aa052518ed5eb9d17f237e5c3ba852bde21c46ed111d86
3
+ size 15997
checkpoint-115000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4f9fd7fea9fc9392184113bedd878a0dc2e3453a83445151e62de04095b42ec
3
+ size 15997
checkpoint-115000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:869bef238cbdfebff9c8753957277e82c13ae877ad267b55c467db4f8bccd17c
3
+ size 15997
checkpoint-115000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4639d8473a540f7a3187b718dbbcd86b6af00cc431a475e5d916ecd6807e1a51
3
+ size 15997
checkpoint-115000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62c93d6e72fea66b4f0aec06c4a67e91a65cb8e1bc9519cec94e6ed705980c45
3
+ size 15984
checkpoint-115000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aee58ff4397dd65526c26126e13a2190d6138bfd05299cde1c80862a842b9ee8
3
+ size 15984
checkpoint-115000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6124b54de588a3eeba381b2806222e7df67d4628ee9a5a734b42ad4ea8e295f6
3
+ size 15984
checkpoint-115000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4547c853da7dc1b149dbe2f546f59bc938a80b170668829d7d34a1694a0c347f
3
+ size 15984
checkpoint-115000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:698357ed95ff125b7e88686cfe877f5cb2eb936f83a189de25b9766935740efb
3
+ size 15984
checkpoint-115000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f3ac9a6c07ab1894a0045a618520197bb38cd339411b01bff120059369ea701
3
+ size 15984
checkpoint-115000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6a19cf96074de923d915b2ff1ab230b06cf63fcae2558df79fdddfe146d38e3
3
+ size 15984
checkpoint-115000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29af53437d40a0ca3b510e19cb281d5a1af09673944b0f733adf08166c511bc8
3
+ size 15984
checkpoint-115000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28b197cf4559414df6314d783e8ad1b0b70373fc6a06ab75aed6cfe7efde0c31
3
+ size 1064
checkpoint-115000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c703d14fec6f4c0514dc0b16d9228d3efef3014eac9d62a47ad21f3dc145000
3
+ size 17800398
checkpoint-115000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8114493e98cd1253de0895ffae7bd4e60431880f0cae75e55fd06cca84a782d
3
+ size 5944