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
48bab4c
·
verified ·
1 Parent(s): 2a9d5be

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-29000/config.json +79 -0
  2. checkpoint-29000/generation_config.json +6 -0
  3. checkpoint-29000/model.safetensors +3 -0
  4. checkpoint-29000/optimizer.pt +3 -0
  5. checkpoint-29000/rng_state_0.pth +3 -0
  6. checkpoint-29000/rng_state_1.pth +3 -0
  7. checkpoint-29000/rng_state_10.pth +3 -0
  8. checkpoint-29000/rng_state_11.pth +3 -0
  9. checkpoint-29000/rng_state_12.pth +3 -0
  10. checkpoint-29000/rng_state_13.pth +3 -0
  11. checkpoint-29000/rng_state_14.pth +3 -0
  12. checkpoint-29000/rng_state_15.pth +3 -0
  13. checkpoint-29000/rng_state_2.pth +3 -0
  14. checkpoint-29000/rng_state_3.pth +3 -0
  15. checkpoint-29000/rng_state_4.pth +3 -0
  16. checkpoint-29000/rng_state_5.pth +3 -0
  17. checkpoint-29000/rng_state_6.pth +3 -0
  18. checkpoint-29000/rng_state_7.pth +3 -0
  19. checkpoint-29000/rng_state_8.pth +3 -0
  20. checkpoint-29000/rng_state_9.pth +3 -0
  21. checkpoint-29000/scheduler.pt +3 -0
  22. checkpoint-29000/trainer_state.json +0 -0
  23. checkpoint-29000/training_args.bin +3 -0
  24. checkpoint-30000/config.json +79 -0
  25. checkpoint-30000/generation_config.json +6 -0
  26. checkpoint-30000/model.safetensors +3 -0
  27. checkpoint-30000/optimizer.pt +3 -0
  28. checkpoint-30000/rng_state_0.pth +3 -0
  29. checkpoint-30000/rng_state_1.pth +3 -0
  30. checkpoint-30000/rng_state_10.pth +3 -0
  31. checkpoint-30000/rng_state_11.pth +3 -0
  32. checkpoint-30000/rng_state_12.pth +3 -0
  33. checkpoint-30000/rng_state_13.pth +3 -0
  34. checkpoint-30000/rng_state_14.pth +3 -0
  35. checkpoint-30000/rng_state_15.pth +3 -0
  36. checkpoint-30000/rng_state_2.pth +3 -0
  37. checkpoint-30000/rng_state_3.pth +3 -0
  38. checkpoint-30000/rng_state_4.pth +3 -0
  39. checkpoint-30000/rng_state_5.pth +3 -0
  40. checkpoint-30000/rng_state_6.pth +3 -0
  41. checkpoint-30000/rng_state_7.pth +3 -0
  42. checkpoint-30000/rng_state_8.pth +3 -0
  43. checkpoint-30000/rng_state_9.pth +3 -0
  44. checkpoint-30000/scheduler.pt +3 -0
  45. checkpoint-30000/trainer_state.json +0 -0
  46. checkpoint-30000/training_args.bin +3 -0
checkpoint-29000/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-29000/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-29000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1672bdac77a3ddd0b800ec82e31a6335031acff14da33197605d2ba1bd3e7ff
3
+ size 4539487822
checkpoint-29000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6874c0369116766685d292c7c665e4881e5b9b7393ea06bf29577c3821d7df6
3
+ size 6632345482
checkpoint-29000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6f3c60aa43bd54f7703c4c7ab94b17ae070aa503adbaa31e26b6b1acba32629
3
+ size 15984
checkpoint-29000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e5b6f128d19625d50bbe3166cfd31f8a415350ca24adbd65e6c447a198069ea
3
+ size 15984
checkpoint-29000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:921ee8ea3d4dcc566c5fdd543415a9ecaafeb01a845289fef67447c64dec15f3
3
+ size 15997
checkpoint-29000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6715f6b7e6364ad6b04fbe8589172764ab698ecbe7fa20e80dc5a76f162b39c2
3
+ size 15997
checkpoint-29000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe7500ab992905087e2174aa82a7f2ebe802e0d290bd5b7a21e75244b7dd67b8
3
+ size 15997
checkpoint-29000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05166b6752e945aa60ceab3aabaad0d8d87987369397318eaa5b8f8d09285754
3
+ size 15997
checkpoint-29000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c97f36d56ebda0a03ef1a56b219a0946698b75142c1dc885ddd717cca0eb3e1
3
+ size 15997
checkpoint-29000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a851b764b0d81153d7b5cb4361330e3b01f06705b1fc16f54f32f69481147c1c
3
+ size 15997
checkpoint-29000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7057417f1a155f0cff6e291435ca505f2b25b68de7492c1d9c35962e602eb68
3
+ size 15984
checkpoint-29000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbf8e47d03d00e876632966691c10bca99d8dc781b51246a00eef79a678a5493
3
+ size 15984
checkpoint-29000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fa77fa3bc71b8b69d6317beda931eabdfa2c13c893c10dd14c894e53a5635ad
3
+ size 15984
checkpoint-29000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:823fb3980047c0e6bee1fdf7d2ba2260f77e49e6f7a33ee7ce8033942b614ef5
3
+ size 15984
checkpoint-29000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:baa85062c559047b756622fcd6c13a08d2cd00360c40f5efdb580fbc1005bdc1
3
+ size 15984
checkpoint-29000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d8b864d4a254611644435677292f70001382564bc9ee4364edf8bd48f5051d1
3
+ size 15984
checkpoint-29000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69c4a52bc4b62362a905418de8d4b9e8875706048a079bffbc50a4f3f6da5cd7
3
+ size 15984
checkpoint-29000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23df18db98e98eba7464f7b9f06f276a353d417d0449fc1b3d349643fbef17ea
3
+ size 15984
checkpoint-29000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ade057355db9d37bbbf5abc5fd833a3a4e62dbf6db0f13e4ea4d1e95b8b5cbd4
3
+ size 1064
checkpoint-29000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-29000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12c00d4f129c6e4cb1c5e8e3e5cd5ee7f1a29f459ba7571424fbce988ad267ee
3
+ size 6008
checkpoint-30000/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-30000/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-30000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35d0250ad74fdbf8245e8c29e1f76d274b6fcecba3dfc366b7af3d6a00f1d541
3
+ size 4539487822
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9edc2ed210cd15ef5363e9ed212e5178b0dd215a184fa0392d3763c6424db1c1
3
+ size 6632345482
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de5d35ccbe0262eb3fe2ca56b4b94fbf0f182ce3aea40e30104d656ea8fc36a4
3
+ size 15984
checkpoint-30000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd77f013abdcb688f7bd053d30068c6d385a8c45577dec62aa226180ce910d00
3
+ size 15984
checkpoint-30000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:746ecd3cbcd02cfafd4f6edc86ee03cd41ef2def1f4db7b944bed7ae05a4f634
3
+ size 15997
checkpoint-30000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:265cd21e0ec3128df0104adc17afb1f219d73f515ac2d8d18da2190d1daccb8d
3
+ size 15997
checkpoint-30000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59d8a417689cfac70b003d120603836f03828ff13e3ef0125c0ff93c3914759c
3
+ size 15997
checkpoint-30000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94e514fbb6cfb255e8e8398d19a4b196c1a7e6630c18155e557b883e5f841a20
3
+ size 15997
checkpoint-30000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:838b40797ac2a9e5bc18fe9fa0fcb75ddd3ae30b304456d71e47835b82ba5135
3
+ size 15997
checkpoint-30000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1757acf209fa3d8a9ccc9dcbd20ee91e3ebeacb14482b1c584102b28eff7f2b8
3
+ size 15997
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a8430f0b48f5d2efa387956f756ee85ddb407771def17da6b26b23f9e0231c5
3
+ size 15984
checkpoint-30000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82da0b36815ccd027a1ba28c3cd40935e67ef497adb7b044c5082129bb5617d8
3
+ size 15984
checkpoint-30000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a36b268a16adfb286a250365b80514ab86764338ede7764a2432e0730a388ff4
3
+ size 15984
checkpoint-30000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbaee1b4f9bf573c8f554c52c53a83c27da34ff7521bf6fee08100dd02e53acb
3
+ size 15984
checkpoint-30000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01819d7cbb776323a0db9e222d3215eea5db2acb1f8b7f0984748488067d6c08
3
+ size 15984
checkpoint-30000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f014c434f914d5acbfa7d2251c042dd4396245a1389ed96f4cf37c10ecb18a39
3
+ size 15984
checkpoint-30000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:924948ff8b9fc20b0a07e42b24ae46a9b84f2c96157177837e24bf9e234bf276
3
+ size 15984
checkpoint-30000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c938d43582b619eed99bfb19fda654bda87fbccd3c18dedfcf09e80ca2e323d4
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a4e7b2ed346955b5b87c3fd2ab508efd139e94fd0e7ddb084f3d83700b63e85
3
+ size 1064
checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12c00d4f129c6e4cb1c5e8e3e5cd5ee7f1a29f459ba7571424fbce988ad267ee
3
+ size 6008