Hugging Face's logo

zhiyang1
/
vlm_dc-vae-f32c32-sana-1.1-768_epoch-64-0_group-7_no-self-attn-lora_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
22337b8
·
verified ·
1 Parent(s): 01b8d2a

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-10000/config.json +79 -0
  2. checkpoint-10000/generation_config.json +6 -0
  3. checkpoint-10000/model.safetensors +3 -0
  4. checkpoint-10000/optimizer.pt +3 -0
  5. checkpoint-10000/rng_state_0.pth +3 -0
  6. checkpoint-10000/rng_state_1.pth +3 -0
  7. checkpoint-10000/rng_state_2.pth +3 -0
  8. checkpoint-10000/rng_state_3.pth +3 -0
  9. checkpoint-10000/rng_state_4.pth +3 -0
  10. checkpoint-10000/rng_state_5.pth +3 -0
  11. checkpoint-10000/rng_state_6.pth +3 -0
  12. checkpoint-10000/rng_state_7.pth +3 -0
  13. checkpoint-10000/scheduler.pt +3 -0
  14. checkpoint-10000/trainer_state.json +0 -0
  15. checkpoint-10000/training_args.bin +3 -0
  16. checkpoint-10500/config.json +79 -0
  17. checkpoint-10500/generation_config.json +6 -0
  18. checkpoint-10500/model.safetensors +3 -0
  19. checkpoint-10500/optimizer.pt +3 -0
  20. checkpoint-10500/rng_state_0.pth +3 -0
  21. checkpoint-10500/rng_state_1.pth +3 -0
  22. checkpoint-10500/rng_state_2.pth +3 -0
  23. checkpoint-10500/rng_state_3.pth +3 -0
  24. checkpoint-10500/rng_state_4.pth +3 -0
  25. checkpoint-10500/rng_state_5.pth +3 -0
  26. checkpoint-10500/rng_state_6.pth +3 -0
  27. checkpoint-10500/rng_state_7.pth +3 -0
  28. checkpoint-10500/scheduler.pt +3 -0
  29. checkpoint-10500/trainer_state.json +0 -0
  30. checkpoint-10500/training_args.bin +3 -0
checkpoint-10000/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-10000/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-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ef6653f450b0ab31efa63e6013178911ebf1310df3b3ae52eed83a45a75aa53
3
+ size 4539487822
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ae07f34ed6341d1744864b86c85669e5fce6ea7e84ac6d54aa2856eb970f5ad
3
+ size 6632330442
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c68c237151b7117a3a14cc6214d795daa23f8703aa6827c15e30d5b8b43c664c
3
+ size 15984
checkpoint-10000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b498e30897691b7415440aa0abcc33f7338be7f057d7e79f3e8423f1e3d98a4e
3
+ size 15984
checkpoint-10000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5e9dc37bc2edaf4609c015f039542b98ae90b1e061004cadcc85d8fcdd6dd40
3
+ size 15984
checkpoint-10000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf6159b941d2cc9a95e3fa665f123b2b28041e01dac2eca3d4af6134319fba80
3
+ size 15984
checkpoint-10000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c903300095c9221c208b8140dd711d85c6afce0be92576aebf642a1ccc3221e
3
+ size 15984
checkpoint-10000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a50fe4d63920dd8be152f846d165b877a5a1fe9e3b55f3fdd215f6e9a9ff7d3a
3
+ size 15984
checkpoint-10000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58528e5a3b92dbb65667567186d68439fad749e9a221a9a8c05a5c424de68b36
3
+ size 15984
checkpoint-10000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa993de5abffbee9f7017748271ca8e856e7f435ef90e91d8152a1a7e708adf6
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45e1e9caa8d5e0a6b6164550b191ba0b0c93227138124bd184e22ebbea7e7e37
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9bf8282ccd6ff666018c997d1250e291b7783cc9ac448f42c77d9fff0eb96dd
3
+ size 6072
checkpoint-10500/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-10500/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-10500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b31dc10fac02540ca6714bc75f31a3b673c7248b3fd4888c1e772df35dafd696
3
+ size 4539487822
checkpoint-10500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:632706c799cccf7c442454a1753f4485d24b8f914723db6b53c74a61004ea375
3
+ size 6632345482
checkpoint-10500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9bd3642686e4a8bac35264cdfc8900f4cd300b3a20b0631361382a1b7bb94d9
3
+ size 15984
checkpoint-10500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70ba4ec02045c722a52cdfa870dc12730a244a8b961607e7c524a371c89104e5
3
+ size 15984
checkpoint-10500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56222d8431a9ea3c77bafc5da49eed0f263d1402f46065df557d8ba892e7cb83
3
+ size 15984
checkpoint-10500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4401f91b53f7e6c80e23b4eb925c96021f7439a5a26e0f8a93ec7be8f1607e2e
3
+ size 15984
checkpoint-10500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c90e3ba21c72219cbe2e2550f0757ab92af1431a131135727712d0038d8e2b0
3
+ size 15984
checkpoint-10500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7309ed82d106a5ad1b6eaa7b829fba9a7f96daa74afdb6ec629c365aba8ff680
3
+ size 15984
checkpoint-10500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76d264dda4c8c8ccc43dcd21fbb9a1966e43539671c70752d51e87a54c0a63d9
3
+ size 15984
checkpoint-10500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e3266622a39fe7f7cf0af78c18d42e0e8dcba2db0b0cfae5b8070b8aaa95687
3
+ size 15984
checkpoint-10500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:031704f2145e90f05fdc9a20798c1de794ce81473aad0348f33a40c1f47843fd
3
+ size 1064
checkpoint-10500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f31022dc7e99a752978382562084593faca2a355db43e8e2c7c91cfc0d611c9d
3
+ size 6008