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
6c265ab
·
verified ·
1 Parent(s): ae7485a

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-19500/config.json +79 -0
  2. checkpoint-19500/generation_config.json +6 -0
  3. checkpoint-19500/model.safetensors +3 -0
  4. checkpoint-19500/optimizer.pt +3 -0
  5. checkpoint-19500/rng_state_0.pth +3 -0
  6. checkpoint-19500/rng_state_1.pth +3 -0
  7. checkpoint-19500/rng_state_2.pth +3 -0
  8. checkpoint-19500/rng_state_3.pth +3 -0
  9. checkpoint-19500/rng_state_4.pth +3 -0
  10. checkpoint-19500/rng_state_5.pth +3 -0
  11. checkpoint-19500/rng_state_6.pth +3 -0
  12. checkpoint-19500/rng_state_7.pth +3 -0
  13. checkpoint-19500/scheduler.pt +3 -0
  14. checkpoint-19500/trainer_state.json +0 -0
  15. checkpoint-19500/training_args.bin +3 -0
  16. checkpoint-20000/config.json +79 -0
  17. checkpoint-20000/generation_config.json +6 -0
  18. checkpoint-20000/model.safetensors +3 -0
  19. checkpoint-20000/optimizer.pt +3 -0
  20. checkpoint-20000/rng_state_0.pth +3 -0
  21. checkpoint-20000/rng_state_1.pth +3 -0
  22. checkpoint-20000/rng_state_2.pth +3 -0
  23. checkpoint-20000/rng_state_3.pth +3 -0
  24. checkpoint-20000/rng_state_4.pth +3 -0
  25. checkpoint-20000/rng_state_5.pth +3 -0
  26. checkpoint-20000/rng_state_6.pth +3 -0
  27. checkpoint-20000/rng_state_7.pth +3 -0
  28. checkpoint-20000/scheduler.pt +3 -0
  29. checkpoint-20000/trainer_state.json +0 -0
  30. checkpoint-20000/training_args.bin +3 -0
checkpoint-19500/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-19500/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-19500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f312933c6efd5ea656d322b8d6adaacddd5f131eb508d90d3b521fba61b5e6f
3
+ size 4561722280
checkpoint-19500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e3c0c1da387ae9669e7d1d3fe9d85e8e87fcd7b9b50168e6dedc8edf79ce8359
3
+ size 6632330442
checkpoint-19500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78ea06e9208de843b489cdfb84b48a65ac51792d27a83a749af3ecbbc3b2d6d0
3
+ size 15984
checkpoint-19500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2deaf05e9767518d51999de50456d2d71aaace3b8fda6a4b71aa4ab1b0b80c04
3
+ size 15984
checkpoint-19500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3570e1f85617ed216459228de4a5738ba5eef460a56570ef401e09f21a01691f
3
+ size 15984
checkpoint-19500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42879104282c20277b00908edfb2e6f2291b4cbcc49069046c0184cc430f56e5
3
+ size 15984
checkpoint-19500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0f1b12ddccd0c162712109f7df37f0f13fda9bfc51bfd4c71fffe46cd657419
3
+ size 15984
checkpoint-19500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:813af9b133708c0f068cb5c2ed78dbf1ca2e369d1c11b5bbaffcc5443b56e6b8
3
+ size 15984
checkpoint-19500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e3adbd22b0b441f154225e1d6120bb11b4164541f93a46b6e8ecccb9cd61654
3
+ size 15984
checkpoint-19500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b1ba6634c139c92185fe723008bbf36c697bd0246c2490eea4b2d699f0e5865
3
+ size 15984
checkpoint-19500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8c4f4356bf542da20add0a095d0be55bc56c89026280d08edbcb6cd78f335e6
3
+ size 1064
checkpoint-19500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-19500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d003dbd41434513efd586c2974a92e6b4a893209c887e8bb3434122ab4c53dfd
3
+ size 6072
checkpoint-20000/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-20000/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-20000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a5d23bd0ad400382a1617906d218c2c3a716d6b44862e7d40fc93ea328c3055
3
+ size 4561722280
checkpoint-20000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:742803f6fe6abc487ac20cec01e38ceee3b1553fa404982afd4c296acde71033
3
+ size 6632330442
checkpoint-20000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53b9568ada1cd86672baaa066359b6d3f94b8df07be77ee934eac3200d873f6a
3
+ size 15984
checkpoint-20000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5bd5f7f4e3b7d97fe6972af04b05a112dfd1711e67e4839d393b5715a6102255
3
+ size 15984
checkpoint-20000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ff9b8973e1899528542ad2833899bf048a7dbc48c6ba9bd048e8001cce02dec
3
+ size 15984
checkpoint-20000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f2a22f4dbd1b044863c410943efb8004a7f0573ea8ad9265471f05c15e3a538
3
+ size 15984
checkpoint-20000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab0a2c251c9f165797e68f4d08a776ce14e4c53b5e58f0035ad901736139bdf3
3
+ size 15984
checkpoint-20000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e97a4c83d25a3d333b62fa1743e0959af2c4df01d884513c1ddeb5e6aee86e9
3
+ size 15984
checkpoint-20000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4def92d0fa268928b9a31379879ac42ab681e4322c93ee1be34845637983de9
3
+ size 15984
checkpoint-20000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ca9150ff8c7989a2eb58f995834a892d7681d6ed8e986b8efb7d84dd97074db
3
+ size 15984
checkpoint-20000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3e2e11c20cafdc07bdb97287eb14f1d8b64ee1da506824356f200ecd35984722
3
+ size 1064
checkpoint-20000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-20000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d003dbd41434513efd586c2974a92e6b4a893209c887e8bb3434122ab4c53dfd
3
+ size 6072