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
fd6c5a0
·
verified ·
1 Parent(s): f555f71

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-11000/config.json +79 -0
  2. checkpoint-11000/generation_config.json +6 -0
  3. checkpoint-11000/model.safetensors +3 -0
  4. checkpoint-11000/optimizer.pt +3 -0
  5. checkpoint-11000/rng_state_0.pth +3 -0
  6. checkpoint-11000/rng_state_1.pth +3 -0
  7. checkpoint-11000/rng_state_10.pth +3 -0
  8. checkpoint-11000/rng_state_11.pth +3 -0
  9. checkpoint-11000/rng_state_12.pth +3 -0
  10. checkpoint-11000/rng_state_13.pth +3 -0
  11. checkpoint-11000/rng_state_14.pth +3 -0
  12. checkpoint-11000/rng_state_15.pth +3 -0
  13. checkpoint-11000/rng_state_2.pth +3 -0
  14. checkpoint-11000/rng_state_3.pth +3 -0
  15. checkpoint-11000/rng_state_4.pth +3 -0
  16. checkpoint-11000/rng_state_5.pth +3 -0
  17. checkpoint-11000/rng_state_6.pth +3 -0
  18. checkpoint-11000/rng_state_7.pth +3 -0
  19. checkpoint-11000/rng_state_8.pth +3 -0
  20. checkpoint-11000/rng_state_9.pth +3 -0
  21. checkpoint-11000/scheduler.pt +3 -0
  22. checkpoint-11000/trainer_state.json +0 -0
  23. checkpoint-11000/training_args.bin +3 -0
checkpoint-11000/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-11000/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-11000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ca2fa11e28a1fddf8871ef2146a4378254ba9397430c162f71b3566951a8510
3
+ size 4538452374
checkpoint-11000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8525c2a1f762956cb93379e9c5177c2dfbc37bddb056b9bd10820f8cbad7d1b
3
+ size 6630551906
checkpoint-11000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05277e72df0d484b10c6778d0114876fdb5cebadabfbd5b88af0984bca84629c
3
+ size 15984
checkpoint-11000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43efe5dc9bff279359fcada66285ed5377226c750ecb5a74d2c023a04972cce3
3
+ size 15984
checkpoint-11000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c4c7c717501e0c22a1ee77f75fb6d585e6b5a949c45668cfbfd6012cc16b82e
3
+ size 15997
checkpoint-11000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0273a1786f625723c9b8250ff3b25ac81f8a2437b5d332500b23dd7395a881f8
3
+ size 15997
checkpoint-11000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0836ca4fd6327921728276c81d74544d90242c3516d5397ef652d8a52ae267e1
3
+ size 15997
checkpoint-11000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bd4d55c25b846e982beffe749e038448ec3e8df913ce8557beb4dcd14d2bbcb
3
+ size 15997
checkpoint-11000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e7148f8d6c9d88367cc58363d506129ed0df262468b15475e7f28560025bb31
3
+ size 15997
checkpoint-11000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fcbb9f48570cfbbee9c39481bac262a735f7e957bd7776922ae212888f641242
3
+ size 15997
checkpoint-11000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:742623ac62f360cb736da6decbfe45ace0a77a65c7be4e4eae0100d35d5d26ed
3
+ size 15984
checkpoint-11000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33a6fca2dedf60f2f4be8f7ef857af8bfbf9dffa2884901c00f717dff84fe536
3
+ size 15984
checkpoint-11000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b35baa1c974d96d69ba977214696185deb3624360b77928d988fbea0309b973
3
+ size 15984
checkpoint-11000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68e725f3cd099cf66a728d4fe05497f22e4f43804e392a8f5a3e2d52080c8e94
3
+ size 15984
checkpoint-11000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ceb164f04519fa85a571e94489fe834c1b51ceed30c3bff3159870a775b32d16
3
+ size 15984
checkpoint-11000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:958f77d56c4264db09bcacb6a9277f79691ad00bbfc7cee648c4f17ea101e841
3
+ size 15984
checkpoint-11000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c57f3316c4bf433c78cf5403c11c0c6da69b85b2e11a2cd2035b5e3475fa2f36
3
+ size 15984
checkpoint-11000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e73a9c9411b5e3873715e68ed4428628df12cfe66cc4e41aa9f320a0d8d771f5
3
+ size 15984
checkpoint-11000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd1cced2f57a795c5feb2d24b280b76d1ecbc1af8ac6aa33dbccdeba73689ae7
3
+ size 1064
checkpoint-11000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-11000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93d0b2c8bac6791a62137b191cec9ec78110cd43992b10f4afdd5f39c73c1456
3
+ size 5944