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
7cf425c
·
verified ·
1 Parent(s): acb304d

Upload folder using huggingface_hub

Browse files
Files changed (31) hide show
  1. .gitattributes +2 -0
  2. checkpoint-239000/config.json +79 -0
  3. checkpoint-239000/generation_config.json +6 -0
  4. checkpoint-239000/model.safetensors +3 -0
  5. checkpoint-239000/optimizer.pt +3 -0
  6. checkpoint-239000/rng_state_0.pth +3 -0
  7. checkpoint-239000/rng_state_1.pth +3 -0
  8. checkpoint-239000/rng_state_2.pth +3 -0
  9. checkpoint-239000/rng_state_3.pth +3 -0
  10. checkpoint-239000/rng_state_4.pth +3 -0
  11. checkpoint-239000/rng_state_5.pth +3 -0
  12. checkpoint-239000/rng_state_6.pth +3 -0
  13. checkpoint-239000/rng_state_7.pth +3 -0
  14. checkpoint-239000/scheduler.pt +3 -0
  15. checkpoint-239000/trainer_state.json +3 -0
  16. checkpoint-239000/training_args.bin +3 -0
  17. checkpoint-240000/config.json +79 -0
  18. checkpoint-240000/generation_config.json +6 -0
  19. checkpoint-240000/model.safetensors +3 -0
  20. checkpoint-240000/optimizer.pt +3 -0
  21. checkpoint-240000/rng_state_0.pth +3 -0
  22. checkpoint-240000/rng_state_1.pth +3 -0
  23. checkpoint-240000/rng_state_2.pth +3 -0
  24. checkpoint-240000/rng_state_3.pth +3 -0
  25. checkpoint-240000/rng_state_4.pth +3 -0
  26. checkpoint-240000/rng_state_5.pth +3 -0
  27. checkpoint-240000/rng_state_6.pth +3 -0
  28. checkpoint-240000/rng_state_7.pth +3 -0
  29. checkpoint-240000/scheduler.pt +3 -0
  30. checkpoint-240000/trainer_state.json +3 -0
  31. checkpoint-240000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -76,3 +76,5 @@ checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
76
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
  checkpoint-229000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
  checkpoint-230000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
 
76
  checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
77
  checkpoint-229000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
78
  checkpoint-230000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
79
+ checkpoint-239000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
80
+ checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-239000/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-239000/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-239000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47036db54310c5f9de4d6901364c9dbca5c62dc5be0fd49ea8478e712d151ed2
3
+ size 4411759432
checkpoint-239000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd80ec25aa4a3f1af87b9d0fe3706ec1dc388d6cc276cc5eba7d2145de751626
3
+ size 6332050591
checkpoint-239000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9d0f253e9704f8d597c33638d9dd5763f154123202bfe621eb680cad0e75add
3
+ size 16389
checkpoint-239000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05ce3d982423e24772bfe2706ea5789e13cf5d98c00037607c5cdac0490663d7
3
+ size 16389
checkpoint-239000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db270379cd960128dade13a60b725f62dbcafd9e0c1a0307bed34ad389a9ced3
3
+ size 16389
checkpoint-239000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31aeb9e5a11161d28d388b265fe5f90e2c6d9004718fa20fb804572cf78a6b54
3
+ size 16389
checkpoint-239000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8df8caaa6bba0e522034df11e28c38c8f5ccf1ab1a74d0c4723d9a16648a101
3
+ size 16389
checkpoint-239000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61c5acfe7791c8894c2497e0bf4117e722ddc53000a3143c758355256d61653f
3
+ size 16389
checkpoint-239000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:233f817efcf9f2c9bd6de0bdb095c2c9c1d5a787dc593b5049569d80545d1ca8
3
+ size 16389
checkpoint-239000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50bc4fe182d7d9d3526e62842d82ce999784244f4b109f61880592dae34caf3e
3
+ size 16389
checkpoint-239000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:432595060f7791696bc50a003392c262de2785024eddf44498cf6dbcc800a5bb
3
+ size 1465
checkpoint-239000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32639eb71dcc9fc16bcecac3d19b7c9eba7ea992e1a7547b58aacc81c19ec90f
3
+ size 36784616
checkpoint-239000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:523f4dd79967bd4b82db0c7fd0c1b19db9088f27445a64476a235b6fe990db46
3
+ size 6481
checkpoint-240000/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-240000/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-240000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3eb13c6044afe53a6f720165ed877c20418619d768f5236b848f03c22b6c9225
3
+ size 4411759432
checkpoint-240000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03244a292314f1f5d1483e02275c2c7409dd2db3ea33c108fd964a6387b19b31
3
+ size 6332050591
checkpoint-240000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:814868cbf2acad57d8766ebdc991b05df97cfc70c1fef2e45f32c17680a89be7
3
+ size 16389
checkpoint-240000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eeca20b58e96cc69e93e35670918f91fff33965ab6efbc51dd52b212e7fa3e68
3
+ size 16389
checkpoint-240000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c12c777564a05deef8f39cff376692f4a0ada5248a6eaef7a12be94a0ac9e92c
3
+ size 16389
checkpoint-240000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3977731bb63121440453fbd0dba2fa60c3436d436ee368300927ad73538c297
3
+ size 16389
checkpoint-240000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0743edf8617cd50bd5f61888bd5845dcd1fa72e7045572971961acfc3718a57a
3
+ size 16389
checkpoint-240000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70b9eb6ef2285cd83f80103541f39cd6e43eb96df285e97d1e792a3220ad312f
3
+ size 16389
checkpoint-240000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9edacdd1a5c881ea5d20958254f28e208f2b8fc8c4ded2f88ad83e2b3d7388d1
3
+ size 16389
checkpoint-240000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59a444b1605781a151619c245d7caa0d7a0656b7f59687132571502af5e53517
3
+ size 16389
checkpoint-240000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fcae5be38fa42aae38d26138576d26f63014831276b676c013ad6db1ad26ab5
3
+ size 1465
checkpoint-240000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5f840164a5b307d008b24b47536cdfb4dad1c69b46c7aaafb31d1b6106880e9
3
+ size 36938957
checkpoint-240000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:523f4dd79967bd4b82db0c7fd0c1b19db9088f27445a64476a235b6fe990db46
3
+ size 6481