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 Community
jiuhai commited on
Commit
5c1a615
·
verified ·
1 Parent(s): 25055dc

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-319000/config.json +79 -0
  3. checkpoint-319000/generation_config.json +6 -0
  4. checkpoint-319000/model.safetensors +3 -0
  5. checkpoint-319000/optimizer.pt +3 -0
  6. checkpoint-319000/rng_state_0.pth +3 -0
  7. checkpoint-319000/rng_state_1.pth +3 -0
  8. checkpoint-319000/rng_state_2.pth +3 -0
  9. checkpoint-319000/rng_state_3.pth +3 -0
  10. checkpoint-319000/rng_state_4.pth +3 -0
  11. checkpoint-319000/rng_state_5.pth +3 -0
  12. checkpoint-319000/rng_state_6.pth +3 -0
  13. checkpoint-319000/rng_state_7.pth +3 -0
  14. checkpoint-319000/scheduler.pt +3 -0
  15. checkpoint-319000/trainer_state.json +3 -0
  16. checkpoint-319000/training_args.bin +3 -0
  17. checkpoint-320000/config.json +79 -0
  18. checkpoint-320000/generation_config.json +6 -0
  19. checkpoint-320000/model.safetensors +3 -0
  20. checkpoint-320000/optimizer.pt +3 -0
  21. checkpoint-320000/rng_state_0.pth +3 -0
  22. checkpoint-320000/rng_state_1.pth +3 -0
  23. checkpoint-320000/rng_state_2.pth +3 -0
  24. checkpoint-320000/rng_state_3.pth +3 -0
  25. checkpoint-320000/rng_state_4.pth +3 -0
  26. checkpoint-320000/rng_state_5.pth +3 -0
  27. checkpoint-320000/rng_state_6.pth +3 -0
  28. checkpoint-320000/rng_state_7.pth +3 -0
  29. checkpoint-320000/scheduler.pt +3 -0
  30. checkpoint-320000/trainer_state.json +3 -0
  31. checkpoint-320000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -92,3 +92,5 @@ checkpoint-299000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
92
  checkpoint-300000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
93
  checkpoint-309000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
94
  checkpoint-310000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
92
  checkpoint-300000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
93
  checkpoint-309000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
94
  checkpoint-310000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
95
+ checkpoint-319000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
96
+ checkpoint-320000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-319000/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-319000/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-319000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d75e2aa6deb35908a1bbae9e75dcb1a4ef71f61df8833be9e140a122bb5a511d
3
+ size 4411759432
checkpoint-319000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37ad335ae3545e5072ff377630f8cc646d669d672cd71a83ae978fb23e8079fe
3
+ size 6332050591
checkpoint-319000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5cbfbcea79b5f423610fe93a556991bd9948bac83a1c40ee12de2e5c307a4ff7
3
+ size 16389
checkpoint-319000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab81757dd55e04c7dada81429d6d3559bd64b53a1e6183ea2f5c8737cb089d76
3
+ size 16389
checkpoint-319000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d84eb15f5e08f2f07e0996bc4ea21f924af251973885eb276e7aab458788728
3
+ size 16389
checkpoint-319000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b3005d68db0bfe221f7357e699901397dda6c8a7b687e8e26fad210ec0db0daf
3
+ size 16389
checkpoint-319000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cec8c022cca3d485504931c0e125ba64fa81152fb01a0f196566df6a05f95ff6
3
+ size 16389
checkpoint-319000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:346c5fd6c7bf2e0d37d71bfb33a5e85341a131cdcdaa4e1c4fd56284b39c62d0
3
+ size 16389
checkpoint-319000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4df2a6bf776e4471334ca772815da07bf6e92b0740395eebed47ca266596c035
3
+ size 16389
checkpoint-319000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:135a21ebf53bb6227c092442043b29e63cc8c0a4546622c55c52913933747456
3
+ size 16389
checkpoint-319000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bf54c7fea5bda818514d9c8433912d99d128caf5b0a3788d6a0389c7cc98daf
3
+ size 1465
checkpoint-319000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60c485f393d70638598b41f03d6ab37770be712586e4053c09455276311f54a2
3
+ size 49131981
checkpoint-319000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be4d3cea97981caa658e02e5f8216e830829395a133e6c8921c075a23a6865f1
3
+ size 6481
checkpoint-320000/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-320000/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-320000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c86add3dd95d410822c0c5c8c513a46648d8124f0ba3fff484f746e4da71508b
3
+ size 4411759432
checkpoint-320000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2d9acf3c316f172315e708b79ab1370ad004fa24ac5955315d5428ad0244960
3
+ size 6332050591
checkpoint-320000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9525a3e9918613fdafe40f1970078c28834e141653bf790062e559767fb140ae
3
+ size 16389
checkpoint-320000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc333dec94d2c7aeb2ac29f9e8ab01c6381a8557e0b3b2d10363b875990edc51
3
+ size 16389
checkpoint-320000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45972a715831f9a86b0a116be4f52b88a74248721b55667739b45deb367141b9
3
+ size 16389
checkpoint-320000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02423de4e7f154cca9f7d0d975a6cb2caf194db2cc85193899b4ec7273d4a572
3
+ size 16389
checkpoint-320000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1d53ffa3b7aee6e6a8bbc5562d87ff272e056875136b015b329dc94be465719
3
+ size 16389
checkpoint-320000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eed9c91be263499c65ff4384ce7793036a0647b7b296bb3891670f54bda516b3
3
+ size 16389
checkpoint-320000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82948be449d47c4f46ad72b4277e4cda6b1bf2f4de87d020e1ef8b37a1b3d4a9
3
+ size 16389
checkpoint-320000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a7c06c6b25ae5cfc00423f36a85bad0fa5f86025a62df925bd2fb60461083d7
3
+ size 16389
checkpoint-320000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7749c6abcd6ea8a09c4a157176826cd4b3b681ce34425f0186517a47ecb0d024
3
+ size 1465
checkpoint-320000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba4425b2c570513113a8dad32a344f947aa1c0c8bea1b8d890c5edf3cc2dd86d
3
+ size 49286341
checkpoint-320000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be4d3cea97981caa658e02e5f8216e830829395a133e6c8921c075a23a6865f1
3
+ size 6481