Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
b28d7c6
·
verified ·
1 Parent(s): 2619cf2

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-66000/config.json +79 -0
  2. checkpoint-66000/generation_config.json +6 -0
  3. checkpoint-66000/model.safetensors +3 -0
  4. checkpoint-66000/optimizer.pt +3 -0
  5. checkpoint-66000/rng_state_0.pth +3 -0
  6. checkpoint-66000/rng_state_1.pth +3 -0
  7. checkpoint-66000/rng_state_10.pth +3 -0
  8. checkpoint-66000/rng_state_11.pth +3 -0
  9. checkpoint-66000/rng_state_12.pth +3 -0
  10. checkpoint-66000/rng_state_13.pth +3 -0
  11. checkpoint-66000/rng_state_14.pth +3 -0
  12. checkpoint-66000/rng_state_15.pth +3 -0
  13. checkpoint-66000/rng_state_2.pth +3 -0
  14. checkpoint-66000/rng_state_3.pth +3 -0
  15. checkpoint-66000/rng_state_4.pth +3 -0
  16. checkpoint-66000/rng_state_5.pth +3 -0
  17. checkpoint-66000/rng_state_6.pth +3 -0
  18. checkpoint-66000/rng_state_7.pth +3 -0
  19. checkpoint-66000/rng_state_8.pth +3 -0
  20. checkpoint-66000/rng_state_9.pth +3 -0
  21. checkpoint-66000/scheduler.pt +3 -0
  22. checkpoint-66000/trainer_state.json +0 -0
  23. checkpoint-66000/training_args.bin +3 -0
checkpoint-66000/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-66000/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-66000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:711cdbc93c0e2b302f862f21dfa334222421fb60898572450c8cceef534fa7d1
3
+ size 4539487822
checkpoint-66000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cabfa506724bcea4be83fe1a544ff39fe1f3fb0cddec3019fdb77e6d39441d0
3
+ size 6632345482
checkpoint-66000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13216555350bd887e6065ef72c71c282f70af53ea120c4ccfc135745fd40f712
3
+ size 15984
checkpoint-66000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fc0c54d856f8a6036c619a758b7da7768c9ceb990875872adaa9b29db128dfd
3
+ size 15984
checkpoint-66000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a7717e219da623c1d3108ccb3a8a13e6224d95efe046c34b3f9010f7b27aebb
3
+ size 15997
checkpoint-66000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d7052b14877cdda1c3e79a6abcb8fa417310635efefdf4147a6b84bdd766666
3
+ size 15997
checkpoint-66000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5136abcec907edc6f62279084c97634478bc8a8e5e993648ee9e8981dada508c
3
+ size 15997
checkpoint-66000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef59ecd71b47bead35ecc8d67916bdb99925bf6311cd22f07750e1d602215c4a
3
+ size 15997
checkpoint-66000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:742c6ddd9616b217df651e96e876621bd30a616ad3a0e9e2a21b6226fd86bed4
3
+ size 15997
checkpoint-66000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60cb17f22a719f043866fab8c0e362e074bb50401609a87dea9cdd1502d66cd0
3
+ size 15997
checkpoint-66000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e744b343872c5a86ba4a8407b30041fea73792e44e30f79353f511760dadc373
3
+ size 15984
checkpoint-66000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68746c56dd4091e7fd3c6b81cca8bd6f2876e5bb017514632eba6d63fc9c8590
3
+ size 15984
checkpoint-66000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:414e2ab88086ed30b69c147faf1eafda364709bb1ca0362a47950194d466b6a7
3
+ size 15984
checkpoint-66000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd8370a8b37c9427d425b4fa7472308e780500c160583cd600868998b2cb0a42
3
+ size 15984
checkpoint-66000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:066332425d83de1f3d0e985738d2582d924032d0a070ebaa99d948b48a9b510d
3
+ size 15984
checkpoint-66000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a2fe84d886d45a7a100426e2796449a241d8b317a723ad9c77b8bd44f264064
3
+ size 15984
checkpoint-66000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edde4627089842f8d5f5ceb811265080016753f37af4e3a391411a9de4519a25
3
+ size 15984
checkpoint-66000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c030f3d4f4998c879f24bb4c485297c3dc0f223c08f104f54cb6c1cab511482
3
+ size 15984
checkpoint-66000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:503b7d895b76f13e13a5f405d48390a5a2bf2cf8ee77e13f8d9a855cbeb44962
3
+ size 1064
checkpoint-66000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-66000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ec282b33f93fc47f75b2e2cd037b83cf61f8804600c7e48226080db637ff73f9
3
+ size 6008