Hugging Face's logo

zhiyang1
/
SZ_multinode_dc-ae-f32c32-1.0-768_baseline

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
b39ce48
·
verified ·
1 Parent(s): 053b309

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-13000/config.json +79 -0
  2. checkpoint-13000/generation_config.json +6 -0
  3. checkpoint-13000/model.safetensors +3 -0
  4. checkpoint-13000/optimizer.pt +3 -0
  5. checkpoint-13000/rng_state_0.pth +3 -0
  6. checkpoint-13000/rng_state_1.pth +3 -0
  7. checkpoint-13000/rng_state_10.pth +3 -0
  8. checkpoint-13000/rng_state_11.pth +3 -0
  9. checkpoint-13000/rng_state_12.pth +3 -0
  10. checkpoint-13000/rng_state_13.pth +3 -0
  11. checkpoint-13000/rng_state_14.pth +3 -0
  12. checkpoint-13000/rng_state_15.pth +3 -0
  13. checkpoint-13000/rng_state_2.pth +3 -0
  14. checkpoint-13000/rng_state_3.pth +3 -0
  15. checkpoint-13000/rng_state_4.pth +3 -0
  16. checkpoint-13000/rng_state_5.pth +3 -0
  17. checkpoint-13000/rng_state_6.pth +3 -0
  18. checkpoint-13000/rng_state_7.pth +3 -0
  19. checkpoint-13000/rng_state_8.pth +3 -0
  20. checkpoint-13000/rng_state_9.pth +3 -0
  21. checkpoint-13000/scheduler.pt +3 -0
  22. checkpoint-13000/trainer_state.json +0 -0
  23. checkpoint-13000/training_args.bin +3 -0
  24. checkpoint-14000/config.json +79 -0
  25. checkpoint-14000/generation_config.json +6 -0
  26. checkpoint-14000/model.safetensors +3 -0
  27. checkpoint-14000/optimizer.pt +3 -0
  28. checkpoint-14000/rng_state_0.pth +3 -0
  29. checkpoint-14000/rng_state_1.pth +3 -0
  30. checkpoint-14000/rng_state_10.pth +3 -0
  31. checkpoint-14000/rng_state_11.pth +3 -0
  32. checkpoint-14000/rng_state_12.pth +3 -0
  33. checkpoint-14000/rng_state_13.pth +3 -0
  34. checkpoint-14000/rng_state_14.pth +3 -0
  35. checkpoint-14000/rng_state_15.pth +3 -0
  36. checkpoint-14000/rng_state_2.pth +3 -0
  37. checkpoint-14000/rng_state_3.pth +3 -0
  38. checkpoint-14000/rng_state_4.pth +3 -0
  39. checkpoint-14000/rng_state_5.pth +3 -0
  40. checkpoint-14000/rng_state_6.pth +3 -0
  41. checkpoint-14000/rng_state_7.pth +3 -0
  42. checkpoint-14000/rng_state_8.pth +3 -0
  43. checkpoint-14000/rng_state_9.pth +3 -0
  44. checkpoint-14000/scheduler.pt +3 -0
  45. checkpoint-14000/trainer_state.json +0 -0
  46. checkpoint-14000/training_args.bin +3 -0
checkpoint-13000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": false,
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-13000/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-13000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9462cc4636e717a33eeaaa66f73482e62cb622b4a45dddb6f48f85ba19e7b77d
3
+ size 4410723984
checkpoint-13000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:799a25c005fac1009447321a298ad8b3c06196a01188b6b1b433ab5b33898081
3
+ size 6330248922
checkpoint-13000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48bc37b12e3119071dbb73bd543fafc0b67a58b1ade42eb415014d36fb73bd15
3
+ size 15984
checkpoint-13000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6724c8d4d12fb0347b69d133e321a30142502c7bf907d06c62504552f16d9d7b
3
+ size 15984
checkpoint-13000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02e26be8f6929657579cb636f82048202dff4c5661ffaf2a260d41700925719a
3
+ size 15997
checkpoint-13000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d6a8ce2ac9e29668582964972e211ccc84e0e0ef688eb42e53ff93341c1b7ac
3
+ size 15997
checkpoint-13000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:97639ef4eec07d5a1d6a138ddd7db4d56daed83d314e0f9b973ef3217c78dacc
3
+ size 15997
checkpoint-13000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b383ef1fe0b9a2191d1c7918ce4eb32cac539a625bac39ce10a3232bda5989c6
3
+ size 15997
checkpoint-13000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fc7be38d5bed0728d658a14faecd090dec3224e7e3b9f9d72489c6ca801960e
3
+ size 15997
checkpoint-13000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ff57d5845bb1eaf2c088894c87efc6b2837713dd9b172d2bb2dc757dba00f09
3
+ size 15997
checkpoint-13000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:459f19c393cbf122da097bc53e48aac83b9861ca1f4a2464c0c705f535ef6a0f
3
+ size 15984
checkpoint-13000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1ab5a371549f51dcb5c7b14f267aeb91f8fa4e3c5451ed313c2f7da62189beb
3
+ size 15984
checkpoint-13000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff759c95893d9229e7212f2f1db6b49442b6513892216b6724a96ef212514976
3
+ size 15984
checkpoint-13000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cad84511b7d25b62ad6066979ef13a081026d06863c8633cf3887a2a94e9ccf3
3
+ size 15984
checkpoint-13000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1dadc65e1c2ae3811342ec9e97c1bf01e3a8ad38243cf88020e7c758d9a08d20
3
+ size 15984
checkpoint-13000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd690d97e009e0c8d57a49ef0a0de46519c902cdebc7ae37523b0a40f00c8156
3
+ size 15984
checkpoint-13000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:008e53c470cb0ad1daaa24f9ba03214b6ae0a1f1514d03d01b0db46bdc79f88f
3
+ size 15984
checkpoint-13000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8dc9f43f44993e8c6e6c1ee08711cb3a259c1076203d8aacf3c226ba7e77877
3
+ size 15984
checkpoint-13000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b0e3d20a60f4eafaad713f9b8d678e50d28b2881b87cd3bac4f9b47a2f6fc77
3
+ size 1064
checkpoint-13000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-13000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4619b0f63f4a7c32fb1f5c76861118306be06fb3616ef3d3b50b2fbd9b4d7283
3
+ size 5944
checkpoint-14000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": false,
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-14000/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-14000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db62f234d17ca7f240bbd309fb8d2f5a442f4acb49bdeba450434abf5471476f
3
+ size 4410723984
checkpoint-14000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5ec70953a9308b5ce02ccc27614d9e8620140c3f3be5224aaf588a12956fb42
3
+ size 6330255386
checkpoint-14000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e669aa1ad58ea40ad014e628b1afcaa081b2c112ff76f404fd9adda03004a16c
3
+ size 15984
checkpoint-14000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22b29239e02f84fbe431887af70851d2b821d58a757bd7a0ce41830d175b6d9f
3
+ size 15984
checkpoint-14000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e008f32be6c2b2a90e6bef822e073fe4fe67ee834ab43c1597e6cc4db28b7661
3
+ size 15997
checkpoint-14000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb970a31fb8a20cdd47ab6dbe033d8a18f36f50bf9cfabca7f0d2c1742a9e51a
3
+ size 15997
checkpoint-14000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bcf91ab3f9459ed0ee7ba90499b65f7ef2886e54e2ea9fc55eaa525db21e7519
3
+ size 15997
checkpoint-14000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33a5483736069be1833d233cfdbd85f7cb72f5f7191e9ca0a5876668e8a6fbf0
3
+ size 15997
checkpoint-14000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:57a1d90727f64f8032206101446fc1414a9b078496ed8b7791ccc9579a6272d6
3
+ size 15997
checkpoint-14000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c3b47d739d8c333d7bba4511bef2e655a7654ab280130b98b0fe6fa473f33a9
3
+ size 15997
checkpoint-14000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1d289dba59dc4daa4f2930aef1ef6598190ae38414bb85c4a6c6957e520d301
3
+ size 15984
checkpoint-14000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87a79c9d7e15b1a2535bc7a89baa20703ee3a045160043f0f4b58b6b974200a1
3
+ size 15984
checkpoint-14000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81f9a5500fb0c54989f699b66b0d9bd7a1a1cba1a622745ef6c1291001bfad80
3
+ size 15984
checkpoint-14000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9719917ec4c7be3c3fe26f00eed0a995242c68cee17cfc358e8f33954efa3500
3
+ size 15984
checkpoint-14000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2cf567c2da95cfead9ae527345c0b1672803df8d55eba635ed47ffe164707e29
3
+ size 15984
checkpoint-14000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb7dc6809e3a6b0a729054083c69615d0dcf5a4593dbae5f5693c0789f2f4a27
3
+ size 15984
checkpoint-14000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb85d44a45465a4f80da15e380a13859f01ee58be2f6edfb9fce6a2fc808eab6
3
+ size 15984
checkpoint-14000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:184d52c2bf29f05c4e4f55e7d6135bbd727d710445b1ecbea537ea04372dc651
3
+ size 15984
checkpoint-14000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36204a320e686511944c4e6f537edbc4d44f42115e9a8d4a4ab821b361aa04fc
3
+ size 1064
checkpoint-14000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-14000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70c243d24a7aca99704e4471c6182a30710942d0416a12a91df43c2711d697aa
3
+ size 5944