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
977d1f1
·
verified ·
1 Parent(s): b0bc7c3

Upload folder using huggingface_hub

Browse files
Files changed (23) hide show
  1. checkpoint-50000/config.json +79 -0
  2. checkpoint-50000/generation_config.json +6 -0
  3. checkpoint-50000/model.safetensors +3 -0
  4. checkpoint-50000/optimizer.pt +3 -0
  5. checkpoint-50000/rng_state_0.pth +3 -0
  6. checkpoint-50000/rng_state_1.pth +3 -0
  7. checkpoint-50000/rng_state_10.pth +3 -0
  8. checkpoint-50000/rng_state_11.pth +3 -0
  9. checkpoint-50000/rng_state_12.pth +3 -0
  10. checkpoint-50000/rng_state_13.pth +3 -0
  11. checkpoint-50000/rng_state_14.pth +3 -0
  12. checkpoint-50000/rng_state_15.pth +3 -0
  13. checkpoint-50000/rng_state_2.pth +3 -0
  14. checkpoint-50000/rng_state_3.pth +3 -0
  15. checkpoint-50000/rng_state_4.pth +3 -0
  16. checkpoint-50000/rng_state_5.pth +3 -0
  17. checkpoint-50000/rng_state_6.pth +3 -0
  18. checkpoint-50000/rng_state_7.pth +3 -0
  19. checkpoint-50000/rng_state_8.pth +3 -0
  20. checkpoint-50000/rng_state_9.pth +3 -0
  21. checkpoint-50000/scheduler.pt +3 -0
  22. checkpoint-50000/trainer_state.json +0 -0
  23. checkpoint-50000/training_args.bin +3 -0
checkpoint-50000/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-50000/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-50000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b4fbb3cac38c2d9f8851d33f3acac2c702712c8e6b744c157c82db67d43afae
3
+ size 4538452374
checkpoint-50000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b17835dd2c135d512a1a3d4b24eadb545b61dca441994dab2f30c051936c286
3
+ size 6630551906
checkpoint-50000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92d4e327f2800ade19fc9bce14fed43bde435169de3e5afc50df091e18bd61c5
3
+ size 15984
checkpoint-50000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af87aa3c604d8053127e65f7b57581431f477fb4de1bb432d58f2f6d24196a31
3
+ size 15984
checkpoint-50000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e3613d748c730437e51227fc24df57ebe72fbdcefe2b6bccdc95d9a642fa520
3
+ size 15997
checkpoint-50000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50928186d69af52549e6f889f0899e1e1ef29613f1b44a84d742cfb48a7a6130
3
+ size 15997
checkpoint-50000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:baced59f887d50068b8ec8d95bca3069ea7cc5c4b2158e1d6393280d7bcce70d
3
+ size 15997
checkpoint-50000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f2c8764793a7e6efb2554176b20ad4a0d41f698bfd81499e34d4bbc5ecbb8db
3
+ size 15997
checkpoint-50000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44a89f71de8d3fcc8c6b7fbbf159cab947563a770ae75e8d8f6cc0250ef405b9
3
+ size 15997
checkpoint-50000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8f2f2a4bb3f46ecd783f97259ae5ee30ec80ffbb22a8c4e8b782a19fe881c9f
3
+ size 15997
checkpoint-50000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:711b5c843285f066b2e287032dfc1fc677c71142112b7dfd9f7a6e12fc22efaa
3
+ size 15984
checkpoint-50000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8c7d9de8aa278b8ca273f61d41bce0810b4061f6f913a39a88061e2c5b30553
3
+ size 15984
checkpoint-50000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70cb19c4dd2df609cd7eb781da8434cd86f7be62cbdbcc93b54180345876aa47
3
+ size 15984
checkpoint-50000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:924d39d0f67dda32cadbcad0e74b0546543ae9f3111f1bb0f02c56e946d98da9
3
+ size 15984
checkpoint-50000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17be458f3e1e42f13f66f3353813ee746cea1b7e272b296418395fc31314731a
3
+ size 15984
checkpoint-50000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b3abc9f56724cdb456d41a0069dd2993f1dee71975e779e427d84762bd78e1e
3
+ size 15984
checkpoint-50000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:918c7e024d3565f57b6a007614605c11d250f7bc85036d307bdc9107efeb8222
3
+ size 15984
checkpoint-50000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8dfd4c50742db223b8474de6b0da3a82d88599ae6ee3c9a635b8cbabba04bb9
3
+ size 15984
checkpoint-50000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1e8c62b290c5c06f78a386c6d256c09d0705fd53312a2e572aff46a3515e1dbb
3
+ size 1064
checkpoint-50000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-50000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca325fc597c50057582b7d9952aa2d1aac2ecf8e3208a6ae90b4f743ab9ceca7
3
+ size 6008