Hugging Face's logo

zhiyang1
/
vlm_dc-ae-f32c32-in-1.0-diffusers-768_patch-2_epoch-64-0_group-7_5e4_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
484bce2
·
verified ·
1 Parent(s): f1de82d

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-71000/config.json +79 -0
  3. checkpoint-71000/generation_config.json +6 -0
  4. checkpoint-71000/model.safetensors +3 -0
  5. checkpoint-71000/optimizer.pt +3 -0
  6. checkpoint-71000/rng_state_0.pth +3 -0
  7. checkpoint-71000/rng_state_1.pth +3 -0
  8. checkpoint-71000/rng_state_2.pth +3 -0
  9. checkpoint-71000/rng_state_3.pth +3 -0
  10. checkpoint-71000/rng_state_4.pth +3 -0
  11. checkpoint-71000/rng_state_5.pth +3 -0
  12. checkpoint-71000/rng_state_6.pth +3 -0
  13. checkpoint-71000/rng_state_7.pth +3 -0
  14. checkpoint-71000/scheduler.pt +3 -0
  15. checkpoint-71000/trainer_state.json +3 -0
  16. checkpoint-71000/training_args.bin +3 -0
  17. checkpoint-71500/config.json +79 -0
  18. checkpoint-71500/generation_config.json +6 -0
  19. checkpoint-71500/model.safetensors +3 -0
  20. checkpoint-71500/optimizer.pt +3 -0
  21. checkpoint-71500/rng_state_0.pth +3 -0
  22. checkpoint-71500/rng_state_1.pth +3 -0
  23. checkpoint-71500/rng_state_2.pth +3 -0
  24. checkpoint-71500/rng_state_3.pth +3 -0
  25. checkpoint-71500/rng_state_4.pth +3 -0
  26. checkpoint-71500/rng_state_5.pth +3 -0
  27. checkpoint-71500/rng_state_6.pth +3 -0
  28. checkpoint-71500/rng_state_7.pth +3 -0
  29. checkpoint-71500/scheduler.pt +3 -0
  30. checkpoint-71500/trainer_state.json +3 -0
  31. checkpoint-71500/training_args.bin +3 -0
.gitattributes CHANGED
@@ -35,3 +35,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
+ checkpoint-71000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
+ checkpoint-71500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-71000/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-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-71000/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-71000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2671b1e2f57372c2b532b6039f0912335aff28c7206654356e085f20381cea45
3
+ size 4561722280
checkpoint-71000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21aaeb3fdbc67c9df14bc6cf7f9fda8dad1c24cfe7cb8cc3433bbb29ce9f90dd
3
+ size 6632330442
checkpoint-71000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f5e64e26c63163648f8dfc757bd1f2a9df3f0e4b18f8a813b90bf8c7a7f0b36
3
+ size 15984
checkpoint-71000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4d7e976bb8e484c420925d875dcc67018962e2a92b435c68ede6e92eb6b90a1
3
+ size 15984
checkpoint-71000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef00d8fc8d887ed8eb8b5113ef5fcccf93ba7eef4e3ef2ac7aa3fbe5c254201f
3
+ size 15984
checkpoint-71000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1ac70b96333e682d177c95ef5d13d7733e0bc0d78142bbd3fcff5f14a22e5cb
3
+ size 15984
checkpoint-71000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:99d807748c8e6a62e2823d99fe3d7fbfb30131d74b14b7003fe20b94b8fe9608
3
+ size 15984
checkpoint-71000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05c36c4764e4a3b69c25bbfe0343e6b71b90daa427a37616f220ecadf26d6290
3
+ size 15984
checkpoint-71000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9677ea7b23379ad299dd4dd7a6a5edbc17efe4bdb8085bbafd4224094a06b13a
3
+ size 15984
checkpoint-71000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81a5db1d62e8a8898ba198293d835b008713ea1b9ca6b6355bcb87e2d7736633
3
+ size 15984
checkpoint-71000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5ccac4ded6fd613e8aac725e363c899407026f77b78cb40c6d5965d6a1cb28d
3
+ size 1064
checkpoint-71000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6585477cccd24d5bb41245774d627c2788f94e34ceb0bf9d8d66eb073f1840da
3
+ size 10891226
checkpoint-71000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2555c0698f5652146f25df969c8e89bd04a1ccf18288fd738dd0785fc8100965
3
+ size 6072
checkpoint-71500/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-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-71500/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-71500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe76a86fe11ac723068815253479f34c39bab700ab9b110456c9665b494b222b
3
+ size 4561722280
checkpoint-71500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:369516cb08e3728b53333358198695fa63b2573ad2b04d31b189ff7aac79ef68
3
+ size 6632345482
checkpoint-71500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba02e5d8e075631c93cc886e56f397c16b1a5393237db02e3625d57df07a5589
3
+ size 15984
checkpoint-71500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8242454e1e7b4bd47529c4ec0e8247c533b84a466043ced21c3d81492ca25d23
3
+ size 15984
checkpoint-71500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b7261df16adabddfaa56087ea9a14330ec2182fc1934a4791c483c7935fb7eb
3
+ size 15984
checkpoint-71500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:104ad288889a07e0a4cc47cbb4cedf92bed80a3e55933606c7e65e5b42541262
3
+ size 15984
checkpoint-71500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88dda70d06d9b917c9369e49710b58d12afbe85f681a5763aaa26580e18d1f53
3
+ size 15984
checkpoint-71500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc6e66c1b708fbf61a595a962c90cc9c6d51206f1ceb68fb735eae78f7cd91e8
3
+ size 15984
checkpoint-71500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3b3fc1fd5a5b5892fbd64a933eb7a7eb6ad54f1530057c4805881c1eacd6802
3
+ size 15984
checkpoint-71500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2358bbed997925ca87d1cb897889c01c75feaeb6a9a70f7a3b008c4a77efc36c
3
+ size 15984
checkpoint-71500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:198e6f16ba653686549f76c7dc6a6d5ae8272b7767dfebcf4ea189d5498852f1
3
+ size 1064
checkpoint-71500/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a84a3fc2a778afdbe73d1a2ffc2c740c9fd823e0c1ed803d95a9fe05bfc926f
3
+ size 10968079
checkpoint-71500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1c69e21d1c1f6a6a51b66e20cc10a01e3a34e8a229bdb1e3d1c5a3fc9783acd
3
+ size 6008