Hugging Face's logo

zhiyang1
/
vlm_dc-ae-f32c32-in-1.0-diffusers-768_patch-2_epoch-64-0_group-7_5e4_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
5f34aaa
·
verified ·
1 Parent(s): 24ed8cc

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-39500/config.json +79 -0
  2. checkpoint-39500/generation_config.json +6 -0
  3. checkpoint-39500/model.safetensors +3 -0
  4. checkpoint-39500/optimizer.pt +3 -0
  5. checkpoint-39500/rng_state_0.pth +3 -0
  6. checkpoint-39500/rng_state_1.pth +3 -0
  7. checkpoint-39500/rng_state_2.pth +3 -0
  8. checkpoint-39500/rng_state_3.pth +3 -0
  9. checkpoint-39500/rng_state_4.pth +3 -0
  10. checkpoint-39500/rng_state_5.pth +3 -0
  11. checkpoint-39500/rng_state_6.pth +3 -0
  12. checkpoint-39500/rng_state_7.pth +3 -0
  13. checkpoint-39500/scheduler.pt +3 -0
  14. checkpoint-39500/trainer_state.json +0 -0
  15. checkpoint-39500/training_args.bin +3 -0
  16. checkpoint-40000/config.json +79 -0
  17. checkpoint-40000/generation_config.json +6 -0
  18. checkpoint-40000/model.safetensors +3 -0
  19. checkpoint-40000/optimizer.pt +3 -0
  20. checkpoint-40000/rng_state_0.pth +3 -0
  21. checkpoint-40000/rng_state_1.pth +3 -0
  22. checkpoint-40000/rng_state_2.pth +3 -0
  23. checkpoint-40000/rng_state_3.pth +3 -0
  24. checkpoint-40000/rng_state_4.pth +3 -0
  25. checkpoint-40000/rng_state_5.pth +3 -0
  26. checkpoint-40000/rng_state_6.pth +3 -0
  27. checkpoint-40000/rng_state_7.pth +3 -0
  28. checkpoint-40000/scheduler.pt +3 -0
  29. checkpoint-40000/trainer_state.json +0 -0
  30. checkpoint-40000/training_args.bin +3 -0
checkpoint-39500/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-in-1.0-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-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.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92b19b495a499aa1ffe24ecc7f9bc8d1caafdb9e1f58079cea74194390842455
3
+ size 4561722280
checkpoint-39500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6084e3417af4def572621a42c3eb2c8095bc45247c728caa9ba6ab8e75a41f02
3
+ size 6632330442
checkpoint-39500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e8da3dcc475ed318ea496e07b8482332c7f19998c61862b82a3df6f259ca63b
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:a25a21840909a8e5123c0635e3761a5919a4f1cdee2ffb0661f1f77e5af27844
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:fb3a14086bde47e9418fa9b543737d20809e7d142c8e01e1a6ff37e212d99a0d
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 sha256:7100583f9c14c12180ab1e30210e57d29986206cabf230c38add85aa68f85a43
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 sha256:056df8c8705b702f86842ef46252c747c4c9c17e252d8506b3ea34e0e52d4b00
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 sha256:49a40584e7f7728362719328174226ced666490537e1ce3e347474a8f340510c
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:c3c13acd4fd03e98c9ea03be5658c032baf7fd45fffa48dca5462a14a468de0a
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 sha256:69a22e600b39b26ad7e0ef7f73caca1fb4eeabafabc24b8db50fa03da26160cb
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 sha256:5e96776a8c14cd5190a23c2d455906dacc05b37dd66297e8fb1f7fa72e3b6ef3
3
+ size 6072
checkpoint-40000/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-in-1.0-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-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.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2711eda240ff160a705507195903cdda85084bd1f0a8c311893e10667b9ab5ad
3
+ size 4561722280
checkpoint-40000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c90130133c48acfcd3aca4c8ed9288b4ca6ff10a499cfed9587542dfd76be66
3
+ size 6632330442
checkpoint-40000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73a06fef64cf8e3ea516e44f65c8b29b369be49ee1069b99979660d6cc34a3c8
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 sha256:9c4a771f4a624a62a921668e284fec04c5a4464f20f888a244b0da7ed2dd6f56
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 sha256:287eb7bf931b59f6ff14af25444c4f3703301a3ebbc2d460df6b16e54cf9bb16
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 sha256:2a65cfbbd3f1986a22d0bb06e93cbace616622c6fded6c9e4f7733c6a47a204a
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 sha256:12d0d692b6d15b1bacf223522d9e32fc299a077feaf93d00b32fc0101506b291
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:b16016e0405b162acc22dbd8f219d25746d6fc2cb94a7db7b02d2248557c89f8
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 sha256:3d1cd964b645d3fe25563d0c5135a9ac0eb510126069f6766bf361c679131334
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 sha256:3da5a8f1bd1d653a3d7cf66117d02813ef29de6caa993241b5bbb0d5e25457d3
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 sha256:5e96776a8c14cd5190a23c2d455906dacc05b37dd66297e8fb1f7fa72e3b6ef3
3
+ size 6072