Hugging Face's logo

zhiyang1
/
SZ_multinode_dc-ae-f32c32-1.0-768_baseline

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
59a0cbe
·
verified ·
1 Parent(s): 805c0b6

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-39000/config.json +79 -0
  2. checkpoint-39000/generation_config.json +6 -0
  3. checkpoint-39000/model.safetensors +3 -0
  4. checkpoint-39000/optimizer.pt +3 -0
  5. checkpoint-39000/rng_state_0.pth +3 -0
  6. checkpoint-39000/rng_state_1.pth +3 -0
  7. checkpoint-39000/rng_state_10.pth +3 -0
  8. checkpoint-39000/rng_state_11.pth +3 -0
  9. checkpoint-39000/rng_state_12.pth +3 -0
  10. checkpoint-39000/rng_state_13.pth +3 -0
  11. checkpoint-39000/rng_state_14.pth +3 -0
  12. checkpoint-39000/rng_state_15.pth +3 -0
  13. checkpoint-39000/rng_state_2.pth +3 -0
  14. checkpoint-39000/rng_state_3.pth +3 -0
  15. checkpoint-39000/rng_state_4.pth +3 -0
  16. checkpoint-39000/rng_state_5.pth +3 -0
  17. checkpoint-39000/rng_state_6.pth +3 -0
  18. checkpoint-39000/rng_state_7.pth +3 -0
  19. checkpoint-39000/rng_state_8.pth +3 -0
  20. checkpoint-39000/rng_state_9.pth +3 -0
  21. checkpoint-39000/scheduler.pt +3 -0
  22. checkpoint-39000/trainer_state.json +0 -0
  23. checkpoint-39000/training_args.bin +3 -0
  24. checkpoint-40000/config.json +79 -0
  25. checkpoint-40000/generation_config.json +6 -0
  26. checkpoint-40000/model.safetensors +3 -0
  27. checkpoint-40000/optimizer.pt +3 -0
  28. checkpoint-40000/rng_state_0.pth +3 -0
  29. checkpoint-40000/rng_state_1.pth +3 -0
  30. checkpoint-40000/rng_state_10.pth +3 -0
  31. checkpoint-40000/rng_state_11.pth +3 -0
  32. checkpoint-40000/rng_state_12.pth +3 -0
  33. checkpoint-40000/rng_state_13.pth +3 -0
  34. checkpoint-40000/rng_state_14.pth +3 -0
  35. checkpoint-40000/rng_state_15.pth +3 -0
  36. checkpoint-40000/rng_state_2.pth +3 -0
  37. checkpoint-40000/rng_state_3.pth +3 -0
  38. checkpoint-40000/rng_state_4.pth +3 -0
  39. checkpoint-40000/rng_state_5.pth +3 -0
  40. checkpoint-40000/rng_state_6.pth +3 -0
  41. checkpoint-40000/rng_state_7.pth +3 -0
  42. checkpoint-40000/rng_state_8.pth +3 -0
  43. checkpoint-40000/rng_state_9.pth +3 -0
  44. checkpoint-40000/scheduler.pt +3 -0
  45. checkpoint-40000/trainer_state.json +0 -0
  46. checkpoint-40000/training_args.bin +3 -0
checkpoint-39000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": 1,
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-in-1.0-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-39000/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-39000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b02d6cb939ad7ca4e366c95403092e2493f3ab061ed6901e14dd8b914717ee67
3
+ size 4410723984
checkpoint-39000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:452b5ce844f30c7e753cfd4ad00a5480ce286d4827fbc036050c29138bca42aa
3
+ size 6330255386
checkpoint-39000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84da25956aab67b8d6a569f05996edcd2896e988b57b4e2e7d05273110a257b1
3
+ size 15984
checkpoint-39000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4aaa8357311906b0ec923b5c2d1f282e3c289f8fdbaf2295ed719ee8fae71b13
3
+ size 15984
checkpoint-39000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aedd244eed7da473c1ee95579f6eb6a18986c4e90b3b9f9652406172c9ea585f
3
+ size 15997
checkpoint-39000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:866924cc68a4b51568ee989836affdb72417b8eda7b8975afd767405eee4cb18
3
+ size 15997
checkpoint-39000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d38196c9f6aaf6af53bee40092afa0fd0ae759b5c1ad18ef03899b698bcfa901
3
+ size 15997
checkpoint-39000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48544e509cc3be012a0eb9d30e8e779ba3edef113d3ca47fd00bbcffb31125ba
3
+ size 15997
checkpoint-39000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e25011882cb72aff3e860aca5e4b43eba5cb7e16a6630d736a5b91442d300ef2
3
+ size 15997
checkpoint-39000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65ce4d3f54654de62fbc76fb84dd065f4da839ea1f99f8e1caa5624ae2b7ea01
3
+ size 15997
checkpoint-39000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76b4ab354fa59907307909677894e5bf0060e5cc2023cee97d94c1b4d9e7e830
3
+ size 15984
checkpoint-39000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b4584461e82f480d97cfa164ed394656124dee0a2daae27319de9e910472747
3
+ size 15984
checkpoint-39000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:073acab9a94eb291991db11c66e4b18bceca812418ab7224556a36e25d8bed9f
3
+ size 15984
checkpoint-39000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d34e936848251cdf46e8daad5fc5c90fd8971c6ef32f7c612a2da722a74b2b9
3
+ size 15984
checkpoint-39000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:001c0702e5ce75c78a6616494c60c85ebbe81cc2f8163bf51dd1ee1efb311664
3
+ size 15984
checkpoint-39000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4afc6d0895d7651149eb30c29e08a6c05d4ae41586684b32f263890559cb079
3
+ size 15984
checkpoint-39000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8be6d8f11926907ff835966ae5cb2fef28a956ae08eba74e538fadd4d1f4e188
3
+ size 15984
checkpoint-39000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b069b700d14c6d37f184268644c7fdbc69466b2c181d981709068b74f854a888
3
+ size 15984
checkpoint-39000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51db50218a1516f400252b4633addd27930e9db8194e29d2f3267385fb573eee
3
+ size 1064
checkpoint-39000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e03a831526a5e1ff3ecaad2a58944193fde85d6d18b2730c7c0488d31f3bdc98
3
+ size 5944
checkpoint-40000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": 1,
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-in-1.0-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-40000/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-40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d75d6cff394bc8145c084e8d378eadb28fac59e89afb2cdd9860dd9221b0771d
3
+ size 4410723984
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5edb8c38bff8675f203b07da1a6b4bb72d56686590f84f74a3bf1226b117bd82
3
+ size 6330255386
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:788de595459473d8d86e0a076f56da67b1fd4c75cc3df9b298b9a0c97e734654
3
+ size 15984
checkpoint-40000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2fc89b5898a4d3e635ba2da42302e9a2d8c12cf6829a4688ccc8efb19455951
3
+ size 15984
checkpoint-40000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f90733a9b886a332ce1089ad39ff3e48d1b25510626ea72b39be04f44097961
3
+ size 15997
checkpoint-40000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73a7685ad40bb1492a64b3475c88a470ef60c031157535d2f88fb19400f1576e
3
+ size 15997
checkpoint-40000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:449a96839289e5676e7a1ffbb7a61035607fa3c9ed9e532da895dbc7d9ee12b0
3
+ size 15997
checkpoint-40000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b112c8229e489b964fc3eea4288a17833c97fdb6f37578846182140dc1f16c71
3
+ size 15997
checkpoint-40000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:396276349c485cc21364c9d5f825b16bf50cb425516aeefe2244e2ea652e0c6f
3
+ size 15997
checkpoint-40000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ccfa72d459c63e40d370df1d42b6060b9e220105fd976e45b2938cd2aeaa380
3
+ size 15997
checkpoint-40000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26c4fdb3b2609437e07be4938cef551ea8d73bb18c172a7a465b28d64d8269cd
3
+ size 15984
checkpoint-40000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9749221d6b49cd2986b951b9c6b9218105c3706f64e676e897971e5ad580482
3
+ size 15984
checkpoint-40000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aa945e8a865448ba486f1d0a353577281be46ac0cd4bb2dac3f39f8bc6a8854
3
+ size 15984
checkpoint-40000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5bb21c60b5bb016cff8159e784d7b9680765f6b550e6cd34b1b3945816b6a0bd
3
+ size 15984
checkpoint-40000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a86187561807379b0cadc3fb8fd1824198b5b3486d404c6fb25592b4f6b2721
3
+ size 15984
checkpoint-40000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b6079f6245f745476652ce26c363ab0118a552464e488c468d3f716429546cb
3
+ size 15984
checkpoint-40000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ea3163e3f467144782a8138de34e9cb39b1addb87594261f3cd325fb86cb211
3
+ size 15984
checkpoint-40000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f668179e9d650f30337be9bc36af00534a34b14bf88e919947d116b54bb6a6a5
3
+ size 15984
checkpoint-40000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a6ddc8cb2af1ec91280ca0f9f919ebd0c98af2adfa8a2cbfd6da4e14464b7e1
3
+ size 1064
checkpoint-40000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-40000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e03a831526a5e1ff3ecaad2a58944193fde85d6d18b2730c7c0488d31f3bdc98
3
+ size 5944