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
58a194e
·
verified ·
1 Parent(s): 7f9f2e2

Upload folder using huggingface_hub

Browse files
Files changed (30) hide show
  1. checkpoint-59500/config.json +79 -0
  2. checkpoint-59500/generation_config.json +6 -0
  3. checkpoint-59500/model.safetensors +3 -0
  4. checkpoint-59500/optimizer.pt +3 -0
  5. checkpoint-59500/rng_state_0.pth +3 -0
  6. checkpoint-59500/rng_state_1.pth +3 -0
  7. checkpoint-59500/rng_state_2.pth +3 -0
  8. checkpoint-59500/rng_state_3.pth +3 -0
  9. checkpoint-59500/rng_state_4.pth +3 -0
  10. checkpoint-59500/rng_state_5.pth +3 -0
  11. checkpoint-59500/rng_state_6.pth +3 -0
  12. checkpoint-59500/rng_state_7.pth +3 -0
  13. checkpoint-59500/scheduler.pt +3 -0
  14. checkpoint-59500/trainer_state.json +0 -0
  15. checkpoint-59500/training_args.bin +3 -0
  16. checkpoint-60000/config.json +79 -0
  17. checkpoint-60000/generation_config.json +6 -0
  18. checkpoint-60000/model.safetensors +3 -0
  19. checkpoint-60000/optimizer.pt +3 -0
  20. checkpoint-60000/rng_state_0.pth +3 -0
  21. checkpoint-60000/rng_state_1.pth +3 -0
  22. checkpoint-60000/rng_state_2.pth +3 -0
  23. checkpoint-60000/rng_state_3.pth +3 -0
  24. checkpoint-60000/rng_state_4.pth +3 -0
  25. checkpoint-60000/rng_state_5.pth +3 -0
  26. checkpoint-60000/rng_state_6.pth +3 -0
  27. checkpoint-60000/rng_state_7.pth +3 -0
  28. checkpoint-60000/scheduler.pt +3 -0
  29. checkpoint-60000/trainer_state.json +0 -0
  30. checkpoint-60000/training_args.bin +3 -0
checkpoint-59500/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-59500/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-59500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a59324f1b73232246f0b997951986d08e7ec9a56724e8cd45544de3076fbf9de
3
+ size 4561722280
checkpoint-59500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0d3e28d9613b6a98207d08452686597f8d529f8ec5a5dce2d7beffe585cbf5d
3
+ size 6632330442
checkpoint-59500/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a3320c49a9b245bc96544c0f1762fc6fa6c738b8fe0ebd8df896b218d2ce71b
3
+ size 15984
checkpoint-59500/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:413ab9028fecfad6d924c56ecb545744dd16f3103c29e9e3d530493bfad14438
3
+ size 15984
checkpoint-59500/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b1753224a6034f8d344e198b36c58399bf9c63373a6da30af687c6624402d9c
3
+ size 15984
checkpoint-59500/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1d68680cbad458a084894162cbcd962ca28d1744fe21f5e48ae643ac3784975
3
+ size 15984
checkpoint-59500/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4fac3a12058477cca5677553bef08757c3009b8d2389bb752f2791bbdc931dd0
3
+ size 15984
checkpoint-59500/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:427a39f9996237bbbe519cf1653e9aa1307eea65b3468431c2dfb1b44318fea5
3
+ size 15984
checkpoint-59500/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e149b0c756e3172a097426003b3b6a45deb43e2be8a18fc662846efb3b61b936
3
+ size 15984
checkpoint-59500/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:faa2fe72e9aedc32da38b5d7d7c793a627885d10aa3b1b7a9ddac4fcd1250219
3
+ size 15984
checkpoint-59500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc9304a5805874198d71b3c9741bd9d768879714ef3a8e9f01e3f5c1a4e2631b
3
+ size 1064
checkpoint-59500/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-59500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8fd5da88a7cee19939c1bc189275e2a984e133430703e6d8d0d354e2a7a8467
3
+ size 6072
checkpoint-60000/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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:376ccdecb12953358658b3d7fe84522f5032af848c810fa0f3ab45e4cd9ebf0b
3
+ size 4561722280
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67428c914c12861e0b49275299870b5462ff30b66704f19040af96b89acee681
3
+ size 6632330442
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e56ccaa3a3894eebf1c73ab6af85cfba2503f74341f4ec102c57d7b930d0d076
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:30d833b4cdd7cc42d2c67d220bf30fce3b9f81273fd1c3af19cc895571c2fd9a
3
+ size 15984
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abdc4b6a34e7de279f672ac7bc657ec725a62c077072e79b043bdb6ef5f78b21
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60f752e823d537bf246b858b15cec29c05fdb831d313e26fd62ff480c9852f75
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a024e17eb92b86e43fc07e5cf9dbd26920db677ad38c46ed443f64538e0b81d9
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0d4c094270b4c9f7fdb6114c655747664f35e029e462a1794dd9251ab58ec97
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32faa2f39ee567fdf5705739e350d85487c1ff7ab2dc87ca7adf230f1aa60e12
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03bde4a304ef03d05c450a11168fc31bed04a4e6f70961c1add7ed7d14e28944
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d97e514637e434b2bf16206f070fa73cdd1886ef533977b76474b9e329b4e4c
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8fd5da88a7cee19939c1bc189275e2a984e133430703e6d8d0d354e2a7a8467
3
+ size 6072