Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_aug_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
a623ab2
·
verified ·
1 Parent(s): a421bb4

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-58000/config.json +79 -0
  2. checkpoint-58000/generation_config.json +6 -0
  3. checkpoint-58000/model.safetensors +3 -0
  4. checkpoint-58000/optimizer.pt +3 -0
  5. checkpoint-58000/rng_state_0.pth +3 -0
  6. checkpoint-58000/rng_state_1.pth +3 -0
  7. checkpoint-58000/rng_state_10.pth +3 -0
  8. checkpoint-58000/rng_state_11.pth +3 -0
  9. checkpoint-58000/rng_state_12.pth +3 -0
  10. checkpoint-58000/rng_state_13.pth +3 -0
  11. checkpoint-58000/rng_state_14.pth +3 -0
  12. checkpoint-58000/rng_state_15.pth +3 -0
  13. checkpoint-58000/rng_state_2.pth +3 -0
  14. checkpoint-58000/rng_state_3.pth +3 -0
  15. checkpoint-58000/rng_state_4.pth +3 -0
  16. checkpoint-58000/rng_state_5.pth +3 -0
  17. checkpoint-58000/rng_state_6.pth +3 -0
  18. checkpoint-58000/rng_state_7.pth +3 -0
  19. checkpoint-58000/rng_state_8.pth +3 -0
  20. checkpoint-58000/rng_state_9.pth +3 -0
  21. checkpoint-58000/scheduler.pt +3 -0
  22. checkpoint-58000/trainer_state.json +0 -0
  23. checkpoint-58000/training_args.bin +3 -0
  24. checkpoint-60000/config.json +79 -0
  25. checkpoint-60000/generation_config.json +6 -0
  26. checkpoint-60000/model.safetensors +3 -0
  27. checkpoint-60000/optimizer.pt +3 -0
  28. checkpoint-60000/rng_state_0.pth +3 -0
  29. checkpoint-60000/rng_state_1.pth +3 -0
  30. checkpoint-60000/rng_state_10.pth +3 -0
  31. checkpoint-60000/rng_state_11.pth +3 -0
  32. checkpoint-60000/rng_state_12.pth +3 -0
  33. checkpoint-60000/rng_state_13.pth +3 -0
  34. checkpoint-60000/rng_state_14.pth +3 -0
  35. checkpoint-60000/rng_state_15.pth +3 -0
  36. checkpoint-60000/rng_state_2.pth +3 -0
  37. checkpoint-60000/rng_state_3.pth +3 -0
  38. checkpoint-60000/rng_state_4.pth +3 -0
  39. checkpoint-60000/rng_state_5.pth +3 -0
  40. checkpoint-60000/rng_state_6.pth +3 -0
  41. checkpoint-60000/rng_state_7.pth +3 -0
  42. checkpoint-60000/rng_state_8.pth +3 -0
  43. checkpoint-60000/rng_state_9.pth +3 -0
  44. checkpoint-60000/scheduler.pt +3 -0
  45. checkpoint-60000/trainer_state.json +0 -0
  46. checkpoint-60000/training_args.bin +3 -0
checkpoint-58000/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": 2,
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-sana-1.1-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-58000/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-58000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd3c1160557c39cc682f38da0324ec16e45e805950a1f205f9cd88fa06b134cf
3
+ size 4539487822
checkpoint-58000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89a3607508c7039676040a2550bee3508df58adf8e11edeedd5e1dfd26e07ae6
3
+ size 6632345482
checkpoint-58000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5a4ade62b42aa65b31a855622ee472f49e42d2f1eb941db9ac868387050e093
3
+ size 15984
checkpoint-58000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64ad810678f2149d39dd184b961ad6766f5161d915f011c04d42aff19d05e62e
3
+ size 15984
checkpoint-58000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11c22ce51de84dbd864cda03c975520c163a6342b1467d830359a6db0fe23c10
3
+ size 15997
checkpoint-58000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55220ab2fdbc60ae36198250203d00d4a47f23c2716b4e08ceddca5b609eda7b
3
+ size 15997
checkpoint-58000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cc68f34d45be2a3d4f061fb2170674096d1622bebeab56ba792b0eb7058ae79
3
+ size 15997
checkpoint-58000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eae1cb1a1324e4032390fad07488a43f13774a894b2ecca012271c21d53cc90f
3
+ size 15997
checkpoint-58000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:618fed3429c7eb48b71e6da45fa395e14622a9d110236b18ed1bbd6d209c287d
3
+ size 15997
checkpoint-58000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09d16b43d868681124aa1a9cda56103b2c8eb37ee06e0e0f9cbba796af12f43f
3
+ size 15997
checkpoint-58000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6afb2aa5f6ccf4c4a34d54c0ba0e6b5ff72a66aa5138d6091340f6603e0ad6ca
3
+ size 15984
checkpoint-58000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:896ac2bcd0547a8004703fc3c94b4bbcd7e302d31b30588b31bb185577c42c84
3
+ size 15984
checkpoint-58000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:994c4a77314e21a13075c5c83f3c755d3a49159e030b63d255af93869f062c9c
3
+ size 15984
checkpoint-58000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41ee93f6559c24ba6364c71f1c445bad3ffad22ae1ebac840a4bbe2edd768da7
3
+ size 15984
checkpoint-58000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:717c978a8e1dcd89121ab5f6934c0ee54f50e85a3c9c81ab2644375cbf9ddec4
3
+ size 15984
checkpoint-58000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08cdaf237032e27782b9edc1361a883a2960498efdc6aeba33eda89df95d5c7a
3
+ size 15984
checkpoint-58000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfef9de84b24400789fde23141858b0b132972aae2ec6ee1c5e53e08a920db4b
3
+ size 15984
checkpoint-58000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b7144fce8128615c66d408e7d4c1b81ee6426180ce5931590987a2f65318f5b
3
+ size 15984
checkpoint-58000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56c96c81c37550f13df90fbaa3fac622803ae07ed6d7c9ab18fe10530b3e73d6
3
+ size 1064
checkpoint-58000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-58000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8d909cacf80a769088bbf93048cbde126f7ab9b049afad82a1898e104951dc6
3
+ size 6008
checkpoint-60000/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": 2,
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-sana-1.1-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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e6404dd839bdcb64269a2de407be5a9ddc86e56ac2034389b0054facc64bdf3
3
+ size 4539487822
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb86edb12506490db37f12783aa23883bc6665fa55c378d4824ecf77d913ccca
3
+ size 6632345482
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88bca05f04b7521a2b2f21bbd906c28ae00e22b22c14197eb5fb7a73970b6568
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d7676a1813b016f89b471e0cb1eb2e95e470cd1a57dbc6ffb2b1859ce2dbf6c
3
+ size 15984
checkpoint-60000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e37ff36912383c5377ba65be378995bf2d6a8572e36bb6c0a02daafffad75904
3
+ size 15997
checkpoint-60000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e79efa3d551a88f5a02bb9df79cd7dad8f9dc7ecff753e3ff8d68edf8d965ff
3
+ size 15997
checkpoint-60000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:834bb74d77a3bfa9a5a406b3bf93d1931d48dc76ed372719a9abb2224d25e9d2
3
+ size 15997
checkpoint-60000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:268afff62bc3d98afee605beef399038d8e86af717cb97b841d7e97cd9ca59d8
3
+ size 15997
checkpoint-60000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e96f1f91a1e0a34dbe91c32aac923dc6d905dacaf0610f34dd7ac61091329dfa
3
+ size 15997
checkpoint-60000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dca65d88c4971fa9eec74f1ce759da280d12539e021fdc960308100e9056736
3
+ size 15997
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:529fc48d4f2edf9527c7d0e0c5b5882dab8ff0b8f1a6eabba39c93ee91b28621
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:409f64eeab8555a2f652cdb3d13e42d16c946f2266f72c959b517b24ea08bda2
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd3107f28619acf413c4f4630ba3149f73daef397d98b0e6196e01c6fcd6e099
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36170a195f4f5f323e0198dc94a38377090ebafa568000bb2c3764a178164d87
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95e91005d5802af60a1122e8d4306ac6b57af78dd099483813a5a1f84960dea7
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c351112fe39aedf2acb802753e32c39afe8b1919715b6e346f23fb65c2705f0a
3
+ size 15984
checkpoint-60000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a528753ccc4c4ddcf91976b7b7849a98b90de1a9241b0bdc4065586e384bdb0
3
+ size 15984
checkpoint-60000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a91019c238e864dac0ca0e4ac5b5136ce1d104ba3d6107b3407dbf10da89e88
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d97e514637e434b2bf16206f070fa73cdd1886ef533977b76474b9e329b4e4c
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8d909cacf80a769088bbf93048cbde126f7ab9b049afad82a1898e104951dc6
3
+ size 6008