Hugging Face's logo

zhiyang1
/
vlm_dc-vae-f32c32-sana-1.1-768_epoch-64_combined_t_res_attn_layerwise-0_group-7_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
2dea4e2
·
verified ·
1 Parent(s): 7bbd516

Upload folder using huggingface_hub

Browse files
Files changed (36) hide show
  1. checkpoint-39500/config.json +81 -0
  2. checkpoint-39500/generation_config.json +6 -0
  3. checkpoint-39500/model-00001-of-00003.safetensors +3 -0
  4. checkpoint-39500/model-00002-of-00003.safetensors +3 -0
  5. checkpoint-39500/model-00003-of-00003.safetensors +3 -0
  6. checkpoint-39500/model.safetensors.index.json +0 -0
  7. checkpoint-39500/optimizer.pt +3 -0
  8. checkpoint-39500/rng_state_0.pth +3 -0
  9. checkpoint-39500/rng_state_1.pth +3 -0
  10. checkpoint-39500/rng_state_2.pth +3 -0
  11. checkpoint-39500/rng_state_3.pth +3 -0
  12. checkpoint-39500/rng_state_4.pth +3 -0
  13. checkpoint-39500/rng_state_5.pth +3 -0
  14. checkpoint-39500/rng_state_6.pth +3 -0
  15. checkpoint-39500/rng_state_7.pth +3 -0
  16. checkpoint-39500/scheduler.pt +3 -0
  17. checkpoint-39500/trainer_state.json +0 -0
  18. checkpoint-39500/training_args.bin +3 -0
  19. checkpoint-40000/config.json +81 -0
  20. checkpoint-40000/generation_config.json +6 -0
  21. checkpoint-40000/model-00001-of-00003.safetensors +3 -0
  22. checkpoint-40000/model-00002-of-00003.safetensors +3 -0
  23. checkpoint-40000/model-00003-of-00003.safetensors +3 -0
  24. checkpoint-40000/model.safetensors.index.json +0 -0
  25. checkpoint-40000/optimizer.pt +3 -0
  26. checkpoint-40000/rng_state_0.pth +3 -0
  27. checkpoint-40000/rng_state_1.pth +3 -0
  28. checkpoint-40000/rng_state_2.pth +3 -0
  29. checkpoint-40000/rng_state_3.pth +3 -0
  30. checkpoint-40000/rng_state_4.pth +3 -0
  31. checkpoint-40000/rng_state_5.pth +3 -0
  32. checkpoint-40000/rng_state_6.pth +3 -0
  33. checkpoint-40000/rng_state_7.pth +3 -0
  34. checkpoint-40000/scheduler.pt +3 -0
  35. checkpoint-40000/trainer_state.json +0 -0
  36. checkpoint-40000/training_args.bin +3 -0
checkpoint-39500/config.json ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 6,
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": 256,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 128,
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
+ "text_cross_attention_dim": 2048,
61
+ "text_encoder_name": "Alpha-VLLM/Lumina-Next-SFT-diffusers",
62
+ "tie_word_embeddings": true,
63
+ "torch_dtype": "bfloat16",
64
+ "transformers_version": "4.47.0",
65
+ "use_cache": true,
66
+ "use_repa": false,
67
+ "use_residual_attn": true,
68
+ "use_sliding_window": false,
69
+ "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
70
+ "video_token_id": 151656,
71
+ "vision_config": {
72
+ "hidden_size": 1536,
73
+ "in_chans": 3,
74
+ "model_type": "qwen2_vl",
75
+ "spatial_patch_size": 14
76
+ },
77
+ "vision_end_token_id": 151653,
78
+ "vision_start_token_id": 151652,
79
+ "vision_token_id": 151654,
80
+ "vocab_size": 151936
81
+ }
checkpoint-39500/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-39500/model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad25d022e0209476cf99cdc671c781f391c34f8bb6783c12988a83ca290bd764
3
+ size 4945243384
checkpoint-39500/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab9fcd429e798b2fcbd2934d9cfee355939e9d2325868b7e97cd3772f1ae5475
3
+ size 4987919040
checkpoint-39500/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:191f30ade8f679571b21d13b0a12fe1f8c20e4a1a6dee9aadca3d4b3180da386
3
+ size 97742926
checkpoint-39500/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ed3de2f0198b1ecaaeb9203cc4ec92dd80d07c9c207d3a423de72a5dcec2fbf
3
+ size 7325168852
checkpoint-39500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56d03ecd475b49fa3ba1f2d94baa1757cc9dc8c7d7e1cce083c5bf7a392f672c
3
+ size 15984
checkpoint-39500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bf5c60aa8f22dfeeacc56da28d2c14eab9afd889dd421100489283dcf2f5ea8
3
+ size 15984
checkpoint-39500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3974159f14c5c260254e1cd273b87f590a462e83112969f41881089cb968e9c7
3
+ size 15984
checkpoint-39500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cdca5b93f8110f8765f2bee9d6c3733b66dd23aeefa30f3787695c3957df0e6
3
+ size 15984
checkpoint-39500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fc26edc80d2ca45cedf97a486eca24f4ad54202813af891f48c3bd5f8ba61ba
3
+ size 15984
checkpoint-39500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e2a6ede950787d1fcee3513160a7961cb2f15b3c3f6d7c0bd0a2d274caa40ae
3
+ size 15984
checkpoint-39500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:680c480fb584f197a4c8ca5f3596d975923c042afa75fb855cfca18eb2720387
3
+ size 15984
checkpoint-39500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d91c9e58aaf9539e1dc450813f626a673533d3851ef055aad199069fc0faf31
3
+ size 15984
checkpoint-39500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2272d541c9dd2d06c2761a16d093c276c4469b8f9d4acb59ee605c62f8397d7b
3
+ size 1064
checkpoint-39500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff1a3e7c48b4e544fdfff7dc60dd979d6646b9f694ef139ea3c71bf0e5ee3839
3
+ size 6072
checkpoint-40000/config.json ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 6,
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": 256,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 128,
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
+ "text_cross_attention_dim": 2048,
61
+ "text_encoder_name": "Alpha-VLLM/Lumina-Next-SFT-diffusers",
62
+ "tie_word_embeddings": true,
63
+ "torch_dtype": "bfloat16",
64
+ "transformers_version": "4.47.0",
65
+ "use_cache": true,
66
+ "use_repa": false,
67
+ "use_residual_attn": true,
68
+ "use_sliding_window": false,
69
+ "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
70
+ "video_token_id": 151656,
71
+ "vision_config": {
72
+ "hidden_size": 1536,
73
+ "in_chans": 3,
74
+ "model_type": "qwen2_vl",
75
+ "spatial_patch_size": 14
76
+ },
77
+ "vision_end_token_id": 151653,
78
+ "vision_start_token_id": 151652,
79
+ "vision_token_id": 151654,
80
+ "vocab_size": 151936
81
+ }
checkpoint-40000/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-40000/model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad25d022e0209476cf99cdc671c781f391c34f8bb6783c12988a83ca290bd764
3
+ size 4945243384
checkpoint-40000/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8078324c75d69f842d8543179280f18ed8a3d0e0be1cb5b8aafa759df027418b
3
+ size 4987919040
checkpoint-40000/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:191f30ade8f679571b21d13b0a12fe1f8c20e4a1a6dee9aadca3d4b3180da386
3
+ size 97742926
checkpoint-40000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d336c0d0c910ce3a0173bc510b219c04ef2333b16f99a032f958e558bce203e
3
+ size 7325168852
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f9cb552a9bf5ac565abf3dfc49390cbc08b6a0c3da59ff5c9177d8a2651e9ec
3
+ size 15984
checkpoint-40000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8bcfcb32532c1d31e383c4a3c25a53e6187392a8e6f9b51fde3fdc4e489adeb
3
+ size 15984
checkpoint-40000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51455cd239764b8a594f3383c39637fc6b612760748edf449b48638081b218ca
3
+ size 15984
checkpoint-40000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf973fc5b15caba659cea7327ea7d4b0e361080267cd85a30e0e5a4896d91e41
3
+ size 15984
checkpoint-40000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1a1272180602a2d9ebee8ee50abcfded22b3db1d9bf78c4f577a9a1ad82d847
3
+ size 15984
checkpoint-40000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c624d39a0a6b059c42d10e848be66f1da1fd996785286a395a1debbb97a75bb3
3
+ size 15984
checkpoint-40000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:041a6788489912c4686eeeac58a4ab5c704ccd85dc38a2e5442739c165b61c20
3
+ size 15984
checkpoint-40000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e77a42d9668c53338238355c4fef2ba0acc2b48612aac39245634e068a124b07
3
+ size 15984
checkpoint-40000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:825e1818d8a0f8348d506acdab26ab27594e76e01808880216f9807387891534
3
+ size 1064
checkpoint-40000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-40000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff1a3e7c48b4e544fdfff7dc60dd979d6646b9f694ef139ea3c71bf0e5ee3839
3
+ size 6072