Hugging Face's logo

zhiyang1
/
vlm_dc-ae-f32c32-in-1.0-diffusers-768_patch-2_epoch-64-0_group-7_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
ac0d0ad
·
verified ·
1 Parent(s): 8e982fc

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-79500/config.json +79 -0
  3. checkpoint-79500/generation_config.json +6 -0
  4. checkpoint-79500/model.safetensors +3 -0
  5. checkpoint-79500/optimizer.pt +3 -0
  6. checkpoint-79500/rng_state_0.pth +3 -0
  7. checkpoint-79500/rng_state_1.pth +3 -0
  8. checkpoint-79500/rng_state_2.pth +3 -0
  9. checkpoint-79500/rng_state_3.pth +3 -0
  10. checkpoint-79500/rng_state_4.pth +3 -0
  11. checkpoint-79500/rng_state_5.pth +3 -0
  12. checkpoint-79500/rng_state_6.pth +3 -0
  13. checkpoint-79500/rng_state_7.pth +3 -0
  14. checkpoint-79500/scheduler.pt +3 -0
  15. checkpoint-79500/trainer_state.json +3 -0
  16. checkpoint-79500/training_args.bin +3 -0
  17. checkpoint-80000/config.json +79 -0
  18. checkpoint-80000/generation_config.json +6 -0
  19. checkpoint-80000/model.safetensors +3 -0
  20. checkpoint-80000/optimizer.pt +3 -0
  21. checkpoint-80000/rng_state_0.pth +3 -0
  22. checkpoint-80000/rng_state_1.pth +3 -0
  23. checkpoint-80000/rng_state_2.pth +3 -0
  24. checkpoint-80000/rng_state_3.pth +3 -0
  25. checkpoint-80000/rng_state_4.pth +3 -0
  26. checkpoint-80000/rng_state_5.pth +3 -0
  27. checkpoint-80000/rng_state_6.pth +3 -0
  28. checkpoint-80000/rng_state_7.pth +3 -0
  29. checkpoint-80000/scheduler.pt +3 -0
  30. checkpoint-80000/trainer_state.json +3 -0
  31. checkpoint-80000/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-79500/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
+ checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-79500/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-79500/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-79500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:007a3e6ded1525bd945e143a2a06e5b1e53598372a592378c4f2b59a5ca25b72
3
+ size 4561722280
checkpoint-79500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0bb2318fd7039ab1445d05e389b3605f3523bb65c9ab0910a2408826524f417
3
+ size 6632330442
checkpoint-79500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f663c8e1767b507e022f1c31bab5369a7c3f6010ab4147f1215482791c21651
3
+ size 15984
checkpoint-79500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:374f0a161a260323c37ae9c4d8c1dc807de4460bc9a5cf781c30099c28037589
3
+ size 15984
checkpoint-79500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1da562db03b78d499a95e110634c5e07b82b2494a2f13ac60c86bccf9848d631
3
+ size 15984
checkpoint-79500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22a12c428770b418fe30ee8f691ea6a267a86fc418367d08375595a2c226a0fc
3
+ size 15984
checkpoint-79500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b2ec4bf8818fdc0b4c90a93f76b0f39c29409e7276f76fc4d2efee2420dc8d4
3
+ size 15984
checkpoint-79500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2365ff7c35fb9cc65716b1f3c4c35ab1bd00a20e06a705ad70b51ff6c4660ca
3
+ size 15984
checkpoint-79500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02b666ccceeb1999476edd61b307a3d9570b17fbba87cc27781cd7026a80d41a
3
+ size 15984
checkpoint-79500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fe9606229a1fefafc42ba286f8ceecc19d1c5982823aa68682e1d1260b3f684
3
+ size 15984
checkpoint-79500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28607f6ce1a9d4b8d54e0c3715d628d02d75128894d33111316daa0aa28adf9d
3
+ size 1064
checkpoint-79500/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8608230ea2dd11b14b6c680949f9d05812d5574e4d6094416a9cd2e30b20bf36
3
+ size 12079728
checkpoint-79500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5cd6c67630e1f67b19b6dec095fcc5321298ae9856c4ee90e8d2e14a3ffd5a6
3
+ size 6072
checkpoint-80000/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-80000/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-80000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0c0aa8a9f057aa4a1f0353929b981bc452535057f77f0bfa4a9d2786211d7b4
3
+ size 4561722280
checkpoint-80000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dccf2e8f859b53fa4ee6eab892b28378aa32ea903e6ab38a88abcd31c4905eb7
3
+ size 6632330442
checkpoint-80000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ffca5def89ac3aa00bf5f6b24bf8088eacbd7d5ae0556f9ce514b9a98365e42
3
+ size 15984
checkpoint-80000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c50f8257984a52977a5d71285838e0c94a50c570e9cd0e472c3ab7c847faa99e
3
+ size 15984
checkpoint-80000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d60757e082cf21808dea405a8ec335ab0b30ced15819c4025ecf7460399053c3
3
+ size 15984
checkpoint-80000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:762fdb92084ae1c5c2b8b606d924820a4f18c89687b8ea7f60f6cd0e57b8374d
3
+ size 15984
checkpoint-80000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f02f28b6cb2a88f6185ba43b51e1980cdc291cc8d627002ad6f50fe8b5fd5777
3
+ size 15984
checkpoint-80000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c044bd977132ec745a1a1e9b16e35936d650a20daf02924ddc73a3cc686b573
3
+ size 15984
checkpoint-80000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da5998c84c61bb5d1201e6c860431ae509faf16c13fa404fa4664e1bbd86ef56
3
+ size 15984
checkpoint-80000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ffd3c329ba6519f3e11e1da6ae30c814f3b883d56fa8768b6000d00e21708d3
3
+ size 15984
checkpoint-80000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28ba12d98196a8a695ec1ab1dd2645d9bcc955d9536dba006f419d3d477cfa2b
3
+ size 1064
checkpoint-80000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0dec782025ea5a1751f586f39d039d3d11ecb61fafe7c1684033185053a6e3e5
3
+ size 12155687
checkpoint-80000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5cd6c67630e1f67b19b6dec095fcc5321298ae9856c4ee90e8d2e14a3ffd5a6
3
+ size 6072