Hugging Face's logo

zhiyang1
/
vlm_dc-vae-f32c32-sana-1.1-768_epoch-64_uniform_layerwise-0_group-4_no-self-attn-lora

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
647bade
·
verified ·
1 Parent(s): 5675a60

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-9500/config.json +79 -0
  17. checkpoint-9500/generation_config.json +6 -0
  18. checkpoint-9500/model.safetensors +3 -0
  19. checkpoint-9500/optimizer.pt +3 -0
  20. checkpoint-9500/rng_state_0.pth +3 -0
  21. checkpoint-9500/rng_state_1.pth +3 -0
  22. checkpoint-9500/rng_state_2.pth +3 -0
  23. checkpoint-9500/rng_state_3.pth +3 -0
  24. checkpoint-9500/rng_state_4.pth +3 -0
  25. checkpoint-9500/rng_state_5.pth +3 -0
  26. checkpoint-9500/rng_state_6.pth +3 -0
  27. checkpoint-9500/rng_state_7.pth +3 -0
  28. checkpoint-9500/scheduler.pt +3 -0
  29. checkpoint-9500/trainer_state.json +0 -0
  30. checkpoint-9500/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": 4,
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": false,
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:ef9d18666de5b9057d8576a3ddd7d8f439ddc54e2766c02a3bd7b11dbdf16cf6
3
+ size 4538452374
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa6de62097edb48db80319193ecb0422f733cde2016da067d21d3297c85773ae
3
+ size 6630537250
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a09563688aa3aa79c264b7795e757e72d9fcc581a2f8d242b92b99301e09e533
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 sha256:31531d1247e74ae4224bda193555cc6cd8ddd5c700a74a9bd5f3c96fbc10efcf
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 sha256:2de7eb5eea65333173040f512117fe4f84dfe07b16c5cd7c9be50e006a4c532b
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:ab841d91826856e5862a0038ca8b19f18cd8caf8e04795cce10ad4d082722d15
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 sha256:04802ea9f0f44ffea117b0b7c1eb95ed38fada9dc52b053650fd268f6e9f76d1
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:ea4a8364083a2432c07582f03e6fec28ff8ca58abbda9b62ae56d655c5d2e7b6
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:e2e47e4c13cf5f7baa666304288b9555ae52ce99081c4707870ec107660f001e
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:d707b6b025ad625e3f5871390538c35441675f6bf47a82ab2dce599945ec2177
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cecba43d7459ec5ce9f816c2e7f8c9a6c0974f0e469d7bde6bc33484cab04b5
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:c410ad4c7da9635f6f53c31f6923e5f6692e01a19a302df5c7f17a2639e47423
3
+ size 6072
checkpoint-9500/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": 4,
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": false,
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-9500/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-9500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:901e1cbf6266b1a46564ee00792d5054d8c85d43f0d2788c24c2f514398976c4
3
+ size 4538452374
checkpoint-9500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:578dd393b36ad268507ec5c7bad8e9d8859a0e256c833044c1e689708c63fdea
3
+ size 6630537250
checkpoint-9500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44c54fac7039cbff11244243cd44bc2838945508f66966fa8e800c5ac7e97f59
3
+ size 15984
checkpoint-9500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4f1a196c6fad5a968165b854e154dd24167a94d6277f5665110c48cde7d8dbe
3
+ size 15984
checkpoint-9500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ab238b575f7e142e594a39b059507cb5831ca894dcd2dfa52db8ef0d5632539
3
+ size 15984
checkpoint-9500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba58931434262bc74ef63e03eefd7490517f63b4a800f074a99e20f1585d55b4
3
+ size 15984
checkpoint-9500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c159a16c1873d44a612d1622b6f3e9b94915d2daf4fcde6e100a3765c343f11
3
+ size 15984
checkpoint-9500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48a281b83f9ffa1275523e120616904cf0a2904b62cbe51703edc0579b97f6af
3
+ size 15984
checkpoint-9500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7d4c6817bcaaa47dee56fa6ee17076104bf4d7dcfe1a9500ceedaf2bc8ae646
3
+ size 15984
checkpoint-9500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4bafc89ef07a9f5f9fdb521a1a1d6dd882447de52762bf9c42e7648e1458121a
3
+ size 15984
checkpoint-9500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24954beb59590793d5fc0ddbe7d1a6127ce1ae2ee15502114a025e447199d147
3
+ size 1064
checkpoint-9500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c410ad4c7da9635f6f53c31f6923e5f6692e01a19a302df5c7f17a2639e47423
3
+ size 6072