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
7f1c23f
·
verified ·
1 Parent(s): 12793e2

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-87000/config.json +79 -0
  3. checkpoint-87000/generation_config.json +6 -0
  4. checkpoint-87000/model.safetensors +3 -0
  5. checkpoint-87000/optimizer.pt +3 -0
  6. checkpoint-87000/rng_state_0.pth +3 -0
  7. checkpoint-87000/rng_state_1.pth +3 -0
  8. checkpoint-87000/rng_state_2.pth +3 -0
  9. checkpoint-87000/rng_state_3.pth +3 -0
  10. checkpoint-87000/rng_state_4.pth +3 -0
  11. checkpoint-87000/rng_state_5.pth +3 -0
  12. checkpoint-87000/rng_state_6.pth +3 -0
  13. checkpoint-87000/rng_state_7.pth +3 -0
  14. checkpoint-87000/scheduler.pt +3 -0
  15. checkpoint-87000/trainer_state.json +3 -0
  16. checkpoint-87000/training_args.bin +3 -0
  17. checkpoint-88000/config.json +79 -0
  18. checkpoint-88000/generation_config.json +6 -0
  19. checkpoint-88000/model.safetensors +3 -0
  20. checkpoint-88000/optimizer.pt +3 -0
  21. checkpoint-88000/rng_state_0.pth +3 -0
  22. checkpoint-88000/rng_state_1.pth +3 -0
  23. checkpoint-88000/rng_state_2.pth +3 -0
  24. checkpoint-88000/rng_state_3.pth +3 -0
  25. checkpoint-88000/rng_state_4.pth +3 -0
  26. checkpoint-88000/rng_state_5.pth +3 -0
  27. checkpoint-88000/rng_state_6.pth +3 -0
  28. checkpoint-88000/rng_state_7.pth +3 -0
  29. checkpoint-88000/scheduler.pt +3 -0
  30. checkpoint-88000/trainer_state.json +3 -0
  31. checkpoint-88000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -38,3 +38,5 @@ 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
40
  checkpoint-80000/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
40
  checkpoint-80000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
41
+ checkpoint-87000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
+ checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-87000/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-87000/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-87000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1793e963c01dcb42d52a3035852f4d50da057d6d1e13584bb579e07fead0e27a
3
+ size 4411759432
checkpoint-87000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2598e84a4b5e1f190ce0063a28b37a7e7eb557d8c84fbbdff0bd2f5d7d78e1ff
3
+ size 6332050591
checkpoint-87000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63cc974ab9b41c7d93267a834de9c9fbfd87ba18cd7841b7e60c729b442fdf6e
3
+ size 16389
checkpoint-87000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ccf2cdb295300c309654821dbceb931d7c5e9c2ec38ee6248bb32ad7c4bd739
3
+ size 16389
checkpoint-87000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6ea741595b0cf68a89ea0fc21f71a7293ccf3e61e2ea221bd28593e38fcd0d8
3
+ size 16389
checkpoint-87000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05f74308d5603a7b236877b4e4b5ec0f75cfb3ea3ee835e804f3c0c5dca4ffa3
3
+ size 16389
checkpoint-87000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7ed9dc5e3ae5beda918074c71f216adefa090340ab4cdfde7cd6504584bf0d6
3
+ size 16389
checkpoint-87000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f23907c331361108d9ca7bfb794665cf492cd180d157e6f8dacf8c780f583b12
3
+ size 16389
checkpoint-87000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4ea6658a5bdc25a0d84b874b6a63e59232841ebc746d7d126003223edfa4493
3
+ size 16389
checkpoint-87000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a4fe56396bdcc2ca6a4ce115a82f7f3ae6495a67851ba8a98e8fedce3cf8100
3
+ size 16389
checkpoint-87000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34521f95fbf53ebcb09edd22e43b6bc3e395149380355f54d4ceaeffefb014bf
3
+ size 1465
checkpoint-87000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2d6c54deab762c467ab57efc400a666806052bd8bfe5485fe7e5b09890439c4
3
+ size 13314960
checkpoint-87000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98829495aa335ef9cc8ab2a2eb399cdb6787d3d6143725993d8d956fcb0addd2
3
+ size 6481
checkpoint-88000/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-88000/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-88000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64752f10a3dbac520e4886c4f1fdf97be091aa7ee157f8f321e0a2efd4c1ce8e
3
+ size 4411759432
checkpoint-88000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c53b3c9e987357e8666aec8e5c4e6b399c6d9607952b2d4b771f669bd58b0e66
3
+ size 6332050591
checkpoint-88000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d896eda7b786e4ad93a240d886c35af144c1bb233fd4e71cc83a29d7a202e9a4
3
+ size 16389
checkpoint-88000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e6d50650d7ec98038d46f2744c092d4a1b167ce162e148f07a756e10996e8a4
3
+ size 16389
checkpoint-88000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6074317204f6a84e45f3ab1d7361af5e939992316b227d495233b96631c379c3
3
+ size 16389
checkpoint-88000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3cdcb49c53e6702efdb3fe239f257c19264c7f299a4f3cda8181f6ce98aa1b8
3
+ size 16389
checkpoint-88000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:506ee148bdb48cd58d95c14c9eaf329d197a66fae509e71e7f18e10e224d3186
3
+ size 16389
checkpoint-88000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a433ef7e8376f1b7ef2ea37469fd446fa747da8ef021c5f5a71525232e3856ed
3
+ size 16389
checkpoint-88000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e32c8e469a0a8ccb9a43acf1e847a7de56c6e2bd20260b45487b6b24bc33d9e
3
+ size 16389
checkpoint-88000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf844fda39e8db9b817b77c2a2d0dad10b0f7052e290e42f30c43a02bc44b9e6
3
+ size 16389
checkpoint-88000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e54f7c2a88059e39029ad9fbb05b5ec602a4f3d411a196df7736dab7ad77736a
3
+ size 1465
checkpoint-88000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70ca51ecf1ce1132572ded4dda2ee45b0f726ca6f8bb5e6a936ca07d8b8315f4
3
+ size 13468532
checkpoint-88000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13f4effe07518a7bb63e36fd2b8b914f5d1e516c4ad07ffd34ee8969fae73791
3
+ size 6417