Hugging Face's logo

zhiyang1
/
vlm_dc-vae-f32c32-sana-1.1-768_epoch-64-0_group-14_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
d2e9197
·
verified ·
1 Parent(s): 49891b3

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-19500/config.json +79 -0
  2. checkpoint-19500/generation_config.json +6 -0
  3. checkpoint-19500/model.safetensors +3 -0
  4. checkpoint-19500/optimizer.pt +3 -0
  5. checkpoint-19500/rng_state_0.pth +3 -0
  6. checkpoint-19500/rng_state_1.pth +3 -0
  7. checkpoint-19500/rng_state_2.pth +3 -0
  8. checkpoint-19500/rng_state_3.pth +3 -0
  9. checkpoint-19500/rng_state_4.pth +3 -0
  10. checkpoint-19500/rng_state_5.pth +3 -0
  11. checkpoint-19500/rng_state_6.pth +3 -0
  12. checkpoint-19500/rng_state_7.pth +3 -0
  13. checkpoint-19500/scheduler.pt +3 -0
  14. checkpoint-19500/trainer_state.json +0 -0
  15. checkpoint-19500/training_args.bin +3 -0
  16. checkpoint-20000/config.json +79 -0
  17. checkpoint-20000/generation_config.json +6 -0
  18. checkpoint-20000/model.safetensors +3 -0
  19. checkpoint-20000/optimizer.pt +3 -0
  20. checkpoint-20000/rng_state_0.pth +3 -0
  21. checkpoint-20000/rng_state_1.pth +3 -0
  22. checkpoint-20000/rng_state_2.pth +3 -0
  23. checkpoint-20000/rng_state_3.pth +3 -0
  24. checkpoint-20000/rng_state_4.pth +3 -0
  25. checkpoint-20000/rng_state_5.pth +3 -0
  26. checkpoint-20000/rng_state_6.pth +3 -0
  27. checkpoint-20000/rng_state_7.pth +3 -0
  28. checkpoint-20000/scheduler.pt +3 -0
  29. checkpoint-20000/trainer_state.json +0 -0
  30. checkpoint-20000/training_args.bin +3 -0
checkpoint-19500/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": 14,
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-19500/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-19500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:018c1bf8755e0e580116a701c48847c777c65ed8badc6d513fda126a58e51263
3
+ size 4539487822
checkpoint-19500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:366ea10e1c3a319411fe1d488fde8b36479c586ce6a94f67564e04590ffcff7a
3
+ size 6632479704
checkpoint-19500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecd376fbc95be0b4d42a631a27d798e0c8a19bec23a6d07db6dcfeb2bf8effeb
3
+ size 15984
checkpoint-19500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19aa3f3f493facdc16e38f99f06b1c1b184e6ab82ca4956e66160bee4a1e9707
3
+ size 15984
checkpoint-19500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f9cf87ace4ec13ed1374bb7daad3b924cc243a248da946f773f93b9dee70a22
3
+ size 15984
checkpoint-19500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64df33fa9ebdae3b6aac097abf9d35cda2daf821aeef815e0ed4fe469776b7a0
3
+ size 15984
checkpoint-19500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:832a4c5ae3882247adc7b1723ec4e679944cfe67db8ac200a23b81a37d7dbeca
3
+ size 15984
checkpoint-19500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4819e8100f37a245b296ca4362910a1b80e600c4a62569c5809daf648c0aa3c8
3
+ size 15984
checkpoint-19500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cac3b054f8fc64d1a62e8508d3a4151821c37f0af214009f4030d8051fd47a8
3
+ size 15984
checkpoint-19500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6db2e5b883af20af1aa006e9dfdae73986300f4fec754da63cecdde26b1e7774
3
+ size 15984
checkpoint-19500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8c4f4356bf542da20add0a095d0be55bc56c89026280d08edbcb6cd78f335e6
3
+ size 1064
checkpoint-19500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-19500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be1b47ede3598a7d834a4dd611b80f9637af439ad8f93fa7592c7ec76288e136
3
+ size 6072
checkpoint-20000/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": 14,
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-20000/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-20000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ea62e3b7a6fdd5cfe9b4da285e762cf3d33931b9714c66b9f172161ce3a100a
3
+ size 4539487822
checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad2ea7c6d91d080e79e03795be7926ec35a7030ca4518c9bb8bf59b5bd0e3951
3
+ size 6632479704
checkpoint-20000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02a56212b0e4f66f75402462fe084f2061936ff43743861fb86f5cdc648d3805
3
+ size 15984
checkpoint-20000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e227da46b4d8b2cab52b9463ffe7e8ba166dc65618c6f7c0611febf0527fcb39
3
+ size 15984
checkpoint-20000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bbe532d2e2ce8aac40264c3c3a3764017ce92b8f3e94edb2545008c2217f78f
3
+ size 15984
checkpoint-20000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4854e698d77a7607304d9186c685c803b17a7e2312c78874ca0b21406d03612f
3
+ size 15984
checkpoint-20000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed13f12d7fbd009fa132a7881f8cb6f1473d142ee6453c8791823a0745d742f9
3
+ size 15984
checkpoint-20000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15ec02052bde78dac0ff50249029773a5652daf00747543a7db788a156a5e178
3
+ size 15984
checkpoint-20000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53571ba37eb089ffc75febd655ac675853fd78d1834ee4246f5e54b5d1ca30d1
3
+ size 15984
checkpoint-20000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:878d24b61512edd77e06cf1bd34819075547efd27064faf49ad90ec19ca6ad91
3
+ size 15984
checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e2e11c20cafdc07bdb97287eb14f1d8b64ee1da506824356f200ecd35984722
3
+ size 1064
checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be1b47ede3598a7d834a4dd611b80f9637af439ad8f93fa7592c7ec76288e136
3
+ size 6072