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
8b17540
·
verified ·
1 Parent(s): d2e9197

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-29500/config.json +79 -0
  2. checkpoint-29500/generation_config.json +6 -0
  3. checkpoint-29500/model.safetensors +3 -0
  4. checkpoint-29500/optimizer.pt +3 -0
  5. checkpoint-29500/rng_state_0.pth +3 -0
  6. checkpoint-29500/rng_state_1.pth +3 -0
  7. checkpoint-29500/rng_state_2.pth +3 -0
  8. checkpoint-29500/rng_state_3.pth +3 -0
  9. checkpoint-29500/rng_state_4.pth +3 -0
  10. checkpoint-29500/rng_state_5.pth +3 -0
  11. checkpoint-29500/rng_state_6.pth +3 -0
  12. checkpoint-29500/rng_state_7.pth +3 -0
  13. checkpoint-29500/scheduler.pt +3 -0
  14. checkpoint-29500/trainer_state.json +0 -0
  15. checkpoint-29500/training_args.bin +3 -0
  16. checkpoint-30000/config.json +79 -0
  17. checkpoint-30000/generation_config.json +6 -0
  18. checkpoint-30000/model.safetensors +3 -0
  19. checkpoint-30000/optimizer.pt +3 -0
  20. checkpoint-30000/rng_state_0.pth +3 -0
  21. checkpoint-30000/rng_state_1.pth +3 -0
  22. checkpoint-30000/rng_state_2.pth +3 -0
  23. checkpoint-30000/rng_state_3.pth +3 -0
  24. checkpoint-30000/rng_state_4.pth +3 -0
  25. checkpoint-30000/rng_state_5.pth +3 -0
  26. checkpoint-30000/rng_state_6.pth +3 -0
  27. checkpoint-30000/rng_state_7.pth +3 -0
  28. checkpoint-30000/scheduler.pt +3 -0
  29. checkpoint-30000/trainer_state.json +0 -0
  30. checkpoint-30000/training_args.bin +3 -0
checkpoint-29500/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-29500/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-29500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6556105a7ea16ff46e8dfd8d7e42097a8b93dccc9979eba977bb43e8ce580f3c
3
+ size 4539487822
checkpoint-29500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6673548d075cfa8586355173194c4e87eb563a475708c863ac4f98bd0d4b8269
3
+ size 6632479704
checkpoint-29500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6da2297b227ba66bd6dcdd21ef108fc5f75fdf6aafa056ace9de12d7c6de4039
3
+ size 15984
checkpoint-29500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0345ca95d7854171d241eb7e43c30e62aa260dd72eabd988e1b70cae4141e9e9
3
+ size 15984
checkpoint-29500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5327e0d0ec7b36ab432c0a7a95ad6fbcad4d69309e9db19289b57b05fe6f0f5
3
+ size 15984
checkpoint-29500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f37e2ad39b04090c6463df56f9d37fb1809a713864a6cd3c7a603ab36bc26f3d
3
+ size 15984
checkpoint-29500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a4cf20213f0864a3125c695485c5bef6ab123e5c734b902c11becc1d9b81758
3
+ size 15984
checkpoint-29500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e6dfdb141eb77cc945a198aa4ae835480726a2b16e47143ad3e8d55366a4b12
3
+ size 15984
checkpoint-29500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:910517d24ba2274fe1f910ddcfb5939ce5d038c45ed6aca9c8f4010b32d37f1f
3
+ size 15984
checkpoint-29500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c0ef186cf887e84fa90907f55c2508ce7e9e187ee529ff4958ddcd96fdba921
3
+ size 15984
checkpoint-29500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14627d16fdb6935914f38057a5ffd424a0ccc1aba68d2217f8d12f98b2713205
3
+ size 1064
checkpoint-29500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-29500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db15cac1ac3a4130207df4dd7eddd53e57be0dfb3036d011fd590524cff8e8d7
3
+ size 6072
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": 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-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:d6e707a141629a51a354d90e6b32c9c099f58fe32f3917d8eba611d5afdd1784
3
+ size 4539487822
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb02c0bf709e6ef897c5fc75517a6d0dc9ef019e49e4ee36974d28d0bbdf1e98
3
+ size 6632479704
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:540f082f3a506cda8119734ce3f15c27fe7bf518d8869997d2c6267a7613183e
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 sha256:2b27f99237dbaa99c53c3d745d87093c947c5f4cc6834149797c244a34628375
3
+ size 15984
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3095f3c404ab84c3a5c2f4055aef7ee46e771cd91062afab025c3aa19b490c2a
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 sha256:375a6b83b6b8e47b519fcea9cdf8806d475eeb097c68b942b622d25866fcb5f2
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 sha256:27fb21160d0d175aa482c9793615c304d970ffbd9048486ca0383dfe3eefbf57
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 sha256:952487f5a50a7f8d62c04d8a1f27c933e5eb3a718c7dc58b95be2905a15ef277
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 sha256:858f7421c7e5f6e9d262254c8917019cf0ee2218fcca830a6cdcb551b249d23d
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:b45e67c49e02ed2ae4940c09f050fa7767808d0def1be162039031669432edec
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:db15cac1ac3a4130207df4dd7eddd53e57be0dfb3036d011fd590524cff8e8d7
3
+ size 6072