Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-ae-f32c32-1.0-768_patch-1_group-7

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
d3a2ec4
·
verified ·
1 Parent(s): 13a6f4a

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-22000/config.json +79 -0
  2. checkpoint-22000/generation_config.json +6 -0
  3. checkpoint-22000/model.safetensors +3 -0
  4. checkpoint-22000/optimizer.pt +3 -0
  5. checkpoint-22000/rng_state_0.pth +3 -0
  6. checkpoint-22000/rng_state_1.pth +3 -0
  7. checkpoint-22000/rng_state_10.pth +3 -0
  8. checkpoint-22000/rng_state_11.pth +3 -0
  9. checkpoint-22000/rng_state_12.pth +3 -0
  10. checkpoint-22000/rng_state_13.pth +3 -0
  11. checkpoint-22000/rng_state_14.pth +3 -0
  12. checkpoint-22000/rng_state_15.pth +3 -0
  13. checkpoint-22000/rng_state_2.pth +3 -0
  14. checkpoint-22000/rng_state_3.pth +3 -0
  15. checkpoint-22000/rng_state_4.pth +3 -0
  16. checkpoint-22000/rng_state_5.pth +3 -0
  17. checkpoint-22000/rng_state_6.pth +3 -0
  18. checkpoint-22000/rng_state_7.pth +3 -0
  19. checkpoint-22000/rng_state_8.pth +3 -0
  20. checkpoint-22000/rng_state_9.pth +3 -0
  21. checkpoint-22000/scheduler.pt +3 -0
  22. checkpoint-22000/trainer_state.json +0 -0
  23. checkpoint-22000/training_args.bin +3 -0
  24. checkpoint-23000/config.json +79 -0
  25. checkpoint-23000/generation_config.json +6 -0
  26. checkpoint-23000/model.safetensors +3 -0
  27. checkpoint-23000/optimizer.pt +3 -0
  28. checkpoint-23000/rng_state_0.pth +3 -0
  29. checkpoint-23000/rng_state_1.pth +3 -0
  30. checkpoint-23000/rng_state_10.pth +3 -0
  31. checkpoint-23000/rng_state_11.pth +3 -0
  32. checkpoint-23000/rng_state_12.pth +3 -0
  33. checkpoint-23000/rng_state_13.pth +3 -0
  34. checkpoint-23000/rng_state_14.pth +3 -0
  35. checkpoint-23000/rng_state_15.pth +3 -0
  36. checkpoint-23000/rng_state_2.pth +3 -0
  37. checkpoint-23000/rng_state_3.pth +3 -0
  38. checkpoint-23000/rng_state_4.pth +3 -0
  39. checkpoint-23000/rng_state_5.pth +3 -0
  40. checkpoint-23000/rng_state_6.pth +3 -0
  41. checkpoint-23000/rng_state_7.pth +3 -0
  42. checkpoint-23000/rng_state_8.pth +3 -0
  43. checkpoint-23000/rng_state_9.pth +3 -0
  44. checkpoint-23000/scheduler.pt +3 -0
  45. checkpoint-23000/trainer_state.json +0 -0
  46. checkpoint-23000/training_args.bin +3 -0
checkpoint-22000/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": 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-22000/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-22000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5d948edee93f7cf1ed1b4ba29c9efd7c8f3597a14b87f41d3db96b938f74dc9
3
+ size 4559801712
checkpoint-22000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3c7a50cc8f9b5dbfe82b8add14e3b79a7e2f576af13577166dac262434e042c
3
+ size 6628781666
checkpoint-22000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58e7684a8bbfaf0fd51344e588524cab12617d94d499dd5b33fb22d9f8a56973
3
+ size 15984
checkpoint-22000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b322b39ed27247fcc61ca884771d3e50789c4b3cabe0ca246532f120f2509b47
3
+ size 15984
checkpoint-22000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:546a90580e1bded4f599ce81d58ec684b8a3e1e6a78d403878bdd6620071020f
3
+ size 15997
checkpoint-22000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79898c678295e0c814a5adc9167baa6f0d7a1dae961d71853720ac94c16fe251
3
+ size 15997
checkpoint-22000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef96b82c7d46eca06eff41117ccbc59326e42ffee8c6b69351e9791ad5d62608
3
+ size 15997
checkpoint-22000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab13c56069fba14da1ca6744abddb3432b62bc3c728d5de6dcabc8fbdbb9d3d2
3
+ size 15997
checkpoint-22000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ff1b38cbbe8d19d4b5944b6466788d80ab317e6bb9b6f0ea243f2e3e95cb34b
3
+ size 15997
checkpoint-22000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:471fa02cbf2a6214af7480c5b0458a82ef5ffff53313744f180ba41ba2b84f35
3
+ size 15997
checkpoint-22000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5a8ae0108081d12d205ecd5619d38334b9da88a46e45b9b094c3ab0230db3fd
3
+ size 15984
checkpoint-22000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19de34404efe9f466a769832a2fa2e5040ee5cf6c6a581a74285bedf51e150cb
3
+ size 15984
checkpoint-22000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57571426400c9e4fa34c6875246c6d46fa0e639fb1b34257d36673d0d1d29ffe
3
+ size 15984
checkpoint-22000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f63cca12b5d14dc2c56b2b5d0e643a627be76e0c3d41cb845a2df2262f7aa2a8
3
+ size 15984
checkpoint-22000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:180c8b9652d4b621e7cbe4b476ffd4445738a3f7d11927f931d7b7a5010587b9
3
+ size 15984
checkpoint-22000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:363121759501d9290fa83c51c0cdc179ec53eb68b5a15872052c7b6b149982c1
3
+ size 15984
checkpoint-22000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46bc41efa4f776e26c52b53950a4f7b15d587758e8c5d4d0af97f9c65d58714e
3
+ size 15984
checkpoint-22000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c01f6ce7d98d3dd39994d60190909ecc66fbb63020fb6ffc31cda6495a29b5dd
3
+ size 15984
checkpoint-22000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea9abc1e54f9d68f507eee5a42c1e7074ff6aa349c90f5b697ac94a922a38b0a
3
+ size 1064
checkpoint-22000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-22000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:173f8ee250ced964d43eeb5208ef81d818edd454f95b3d16e74fcb6384a313a0
3
+ size 6008
checkpoint-23000/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": 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-23000/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-23000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8897d4c166c60f6e5c1ef9d6a4ffe387974c83e8cc421ca9486da25b362c1976
3
+ size 4559801712
checkpoint-23000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3e93539929e8fa444f52d8ce706da6239b36340c7658991c2d40d664845f7b6
3
+ size 6628781666
checkpoint-23000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24f7932c621f43e72c4f7719bfbe47e812a7e59ac4c65ef2a50af2b2efe3de41
3
+ size 15984
checkpoint-23000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:069f02a3e718d33cf7a759507787c80c9ecef702ff480f919a409490a28a0f7b
3
+ size 15984
checkpoint-23000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf27c7de7451e795828bbf4f41a7ffbe2fe7ba95f135fe022ab36f0921d322c8
3
+ size 15997
checkpoint-23000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03b14661332caa8ae1780aed0a7858f6a440f477807033a6c78b512cdf9c1f8f
3
+ size 15997
checkpoint-23000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80dd7e5923d4498e9026fff4d367a8b4d9c62a5e749f8cbcf5de9fda44e78cf7
3
+ size 15997
checkpoint-23000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cdf0643e12ac4494c1f51f93e875088b0255c6688ee6f22db8f38c33a9f35f4
3
+ size 15997
checkpoint-23000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50f204977158d9b80777c8ebc1bd54cd384ebd3f1cfd969ede7edc0ef89d824b
3
+ size 15997
checkpoint-23000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca59bc2629e0efaaf607318a827455334e1b00bda6b44866697cc3a79fb56a77
3
+ size 15997
checkpoint-23000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2248652d61de19def479d533a11566a6f4d86d9352ef9bac98d83457d061e430
3
+ size 15984
checkpoint-23000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f01b31001de035145e2569a5a388997c67ce5c831d5ce327e3a8d9c3f378379f
3
+ size 15984
checkpoint-23000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f038581d0df40dde14e26a87ed9a0fd17ba208de41ea71583340670f6e0daf0
3
+ size 15984
checkpoint-23000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:672eda9a0cc71417cb359608c8c5d61bdd61e137d0c96a352c3fc875ae0f4222
3
+ size 15984
checkpoint-23000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d414a7f886aaeb6ff3f13a867a2176ffbe2a8f485d465ba69a855a66c3ae7aa0
3
+ size 15984
checkpoint-23000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69b122e514ab4374c36a72b79fe343218b10299ca22adf2d775e37e596dd276a
3
+ size 15984
checkpoint-23000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1d330ef5469a6520f3f5b157e09425771dd7e1eb0822665afbbefd857aac543
3
+ size 15984
checkpoint-23000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b76d9181e059ed0644eed81013e6a0860f416c9d7585f4ba00bfaa3b93fcb6d
3
+ size 15984
checkpoint-23000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fbaf762795cede88ee3af15172def3bfc86290e95a80dac29e5c573b40493fbc
3
+ size 1064
checkpoint-23000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-23000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f6382c1fd615b0f30351c8d08e695a54c3a3f72743592d38b765211b297d72b
3
+ size 5944