Hugging Face's logo

umd-vt-nyu
/
JH_multinode_dc-ae-f32c32-in-1.0-diffusers-768_patch-1_group-7_5e4_residual_attn

Safetensors
Model card Files
xet
Community
jiuhai commited on
Commit
64c334c
·
verified ·
1 Parent(s): a7b1dce

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-78000/config.json +79 -0
  3. checkpoint-78000/generation_config.json +6 -0
  4. checkpoint-78000/model.safetensors +3 -0
  5. checkpoint-78000/optimizer.pt +3 -0
  6. checkpoint-78000/rng_state_0.pth +3 -0
  7. checkpoint-78000/rng_state_1.pth +3 -0
  8. checkpoint-78000/rng_state_2.pth +3 -0
  9. checkpoint-78000/rng_state_3.pth +3 -0
  10. checkpoint-78000/rng_state_4.pth +3 -0
  11. checkpoint-78000/rng_state_5.pth +3 -0
  12. checkpoint-78000/rng_state_6.pth +3 -0
  13. checkpoint-78000/rng_state_7.pth +3 -0
  14. checkpoint-78000/scheduler.pt +3 -0
  15. checkpoint-78000/trainer_state.json +3 -0
  16. checkpoint-78000/training_args.bin +3 -0
  17. checkpoint-79000/config.json +79 -0
  18. checkpoint-79000/generation_config.json +6 -0
  19. checkpoint-79000/model.safetensors +3 -0
  20. checkpoint-79000/optimizer.pt +3 -0
  21. checkpoint-79000/rng_state_0.pth +3 -0
  22. checkpoint-79000/rng_state_1.pth +3 -0
  23. checkpoint-79000/rng_state_2.pth +3 -0
  24. checkpoint-79000/rng_state_3.pth +3 -0
  25. checkpoint-79000/rng_state_4.pth +3 -0
  26. checkpoint-79000/rng_state_5.pth +3 -0
  27. checkpoint-79000/rng_state_6.pth +3 -0
  28. checkpoint-79000/rng_state_7.pth +3 -0
  29. checkpoint-79000/scheduler.pt +3 -0
  30. checkpoint-79000/trainer_state.json +3 -0
  31. checkpoint-79000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -35,3 +35,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  checkpoint-69000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
37
  checkpoint-70000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
38
+ checkpoint-78000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
39
+ checkpoint-79000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-78000/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": 1,
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-78000/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-78000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4fe89d54015e2e976e63c4406f80df872cef30133c55c5ef2a4c96436fa4b3d
3
+ size 4411759432
checkpoint-78000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19d359f315d3d2ca7f2ef19459e53250af242d56fc266871ce095150149163bc
3
+ size 6332050591
checkpoint-78000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7708331569d54e13c443c1d58e2e5c6abd03871647674ae024ce3e9450eba14
3
+ size 16389
checkpoint-78000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2d4e10c3def2c0d379f207d6e2310cb0c377cef4b0eb2360fd35bc91d7a420b
3
+ size 16389
checkpoint-78000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:485d7554a98d212a99ec3b8b592c929eeedaa1ad04f0cc8ba83763220134fb8f
3
+ size 16389
checkpoint-78000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6b880fdd05fd70ba68765f5b69a8de3cd98e762373ca053a36ced104b8c9562
3
+ size 16389
checkpoint-78000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74d0c21a849eb4fa2984eae63e32d8a392bef8d66ff72e4c19fca9ed9a02ca38
3
+ size 16389
checkpoint-78000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9f3388c1c4b56632c27ba67bb6441c5224cf9467a87ff30051b3f01c1b36acb
3
+ size 16389
checkpoint-78000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:128e49b780e6335916f3916e16823994c0c2efea8b92b2941c1003732654671f
3
+ size 16389
checkpoint-78000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0282ff3cdc0c1e87c111d8f5d0bf56c6ab6591ed8daf0cdf8739cc5d8576f7ae
3
+ size 16389
checkpoint-78000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:437f9d52d794ff5bb6f17bb69f50ee536dfb68f596d6d38b045240f7e219c842
3
+ size 1465
checkpoint-78000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:784738993004cbd8838a45f376b9b7d987c34c39c7700076e07e32db7e3d23ad
3
+ size 11932169
checkpoint-78000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b86f9a0a60355fcc92ce0a31e782836db7d163b33401d30c053d2a914fc12500
3
+ size 6481
checkpoint-79000/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": 1,
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-79000/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-79000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61b5e70e01abd8c5d7e9235ab108f87daa3df890ae0e3c178289b4078e2eb2c3
3
+ size 4411759432
checkpoint-79000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae4fdfa1cb3ec590f706f552e0f03ae03d121684eb6d5db7070be0f081d45b5c
3
+ size 6332050591
checkpoint-79000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:300fc55dd4201e8844b7e09071442d3008a1ab7bd5b612f3281f99d297f7dd0a
3
+ size 16389
checkpoint-79000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4edb3c4956dbb93cef08a9f4c3ce9f15b6a6d06398e075e6f81b257c77d665b4
3
+ size 16389
checkpoint-79000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1488caef6ee93fff10c0aecdbd6211c8175e4e0b879f51ee18752e2ebad2a7de
3
+ size 16389
checkpoint-79000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6980bbefde56dd8301aa846fb9e642ea95df6e81de297bd9075c828c44f77a12
3
+ size 16389
checkpoint-79000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a600b39d01637631c1a853b75c5de96e06aa935892f24e8e74ead2591bf89c1
3
+ size 16389
checkpoint-79000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df0aae57c3a5fea956270c1cccb8a4f6e3de9d9b7846e0ff0f4b068b3aa49ba1
3
+ size 16389
checkpoint-79000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b071617d7ca611535ac76b9c1eb55ed9c4ca74219f7d8ef7681de0044653e5d
3
+ size 16389
checkpoint-79000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6279b6a20c11f5c6678121e63165194fcf44cd215891aa9c9d173a7a10661b44
3
+ size 16389
checkpoint-79000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7209de4509b679e3694fd15c9e9f297953d909ee7d82bf5e2c0dacfc111bfccd
3
+ size 1465
checkpoint-79000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ad5ced38e7c5bf2f26313c6ee89a04b95b0a59d3164bf7c314da514455306ad
3
+ size 12085922
checkpoint-79000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daa05487170aae329a99794e10700eaec57ce32beec374d4922aed232bae17f6
3
+ size 6417