Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
dabf08a
·
verified ·
1 Parent(s): 15bfd39

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-29000/config.json +79 -0
  2. checkpoint-29000/generation_config.json +6 -0
  3. checkpoint-29000/model.safetensors +3 -0
  4. checkpoint-29000/optimizer.pt +3 -0
  5. checkpoint-29000/rng_state_0.pth +3 -0
  6. checkpoint-29000/rng_state_1.pth +3 -0
  7. checkpoint-29000/rng_state_10.pth +3 -0
  8. checkpoint-29000/rng_state_11.pth +3 -0
  9. checkpoint-29000/rng_state_12.pth +3 -0
  10. checkpoint-29000/rng_state_13.pth +3 -0
  11. checkpoint-29000/rng_state_14.pth +3 -0
  12. checkpoint-29000/rng_state_15.pth +3 -0
  13. checkpoint-29000/rng_state_2.pth +3 -0
  14. checkpoint-29000/rng_state_3.pth +3 -0
  15. checkpoint-29000/rng_state_4.pth +3 -0
  16. checkpoint-29000/rng_state_5.pth +3 -0
  17. checkpoint-29000/rng_state_6.pth +3 -0
  18. checkpoint-29000/rng_state_7.pth +3 -0
  19. checkpoint-29000/rng_state_8.pth +3 -0
  20. checkpoint-29000/rng_state_9.pth +3 -0
  21. checkpoint-29000/scheduler.pt +3 -0
  22. checkpoint-29000/trainer_state.json +0 -0
  23. checkpoint-29000/training_args.bin +3 -0
  24. checkpoint-30000/config.json +79 -0
  25. checkpoint-30000/generation_config.json +6 -0
  26. checkpoint-30000/model.safetensors +3 -0
  27. checkpoint-30000/optimizer.pt +3 -0
  28. checkpoint-30000/rng_state_0.pth +3 -0
  29. checkpoint-30000/rng_state_1.pth +3 -0
  30. checkpoint-30000/rng_state_10.pth +3 -0
  31. checkpoint-30000/rng_state_11.pth +3 -0
  32. checkpoint-30000/rng_state_12.pth +3 -0
  33. checkpoint-30000/rng_state_13.pth +3 -0
  34. checkpoint-30000/rng_state_14.pth +3 -0
  35. checkpoint-30000/rng_state_15.pth +3 -0
  36. checkpoint-30000/rng_state_2.pth +3 -0
  37. checkpoint-30000/rng_state_3.pth +3 -0
  38. checkpoint-30000/rng_state_4.pth +3 -0
  39. checkpoint-30000/rng_state_5.pth +3 -0
  40. checkpoint-30000/rng_state_6.pth +3 -0
  41. checkpoint-30000/rng_state_7.pth +3 -0
  42. checkpoint-30000/rng_state_8.pth +3 -0
  43. checkpoint-30000/rng_state_9.pth +3 -0
  44. checkpoint-30000/scheduler.pt +3 -0
  45. checkpoint-30000/trainer_state.json +0 -0
  46. checkpoint-30000/training_args.bin +3 -0
checkpoint-29000/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": false,
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-29000/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-29000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ecdc960940f4620e8bb4efab19fda70a965abb6535f7806c66388423c0ef5da
3
+ size 4538452374
checkpoint-29000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40ed8a3bab44e13b8ef8d3daff1f1e22ae95e329063d71f513483c669c37daca
3
+ size 6630551906
checkpoint-29000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed31fc30bb7913719dd447659b84321e9ff6bae106dd0f9adf0e99895ebfcda7
3
+ size 15984
checkpoint-29000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7049601210c861a8711c5427f1c9ce7c63966982a9ff1fa44f31a36a52bebdfd
3
+ size 15984
checkpoint-29000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:388298b560e93d48b64e239274668a6298df27adfd14c33bb9450661c9462dc6
3
+ size 15997
checkpoint-29000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7097516835174d1fa2c03f4791647bd71522ed218cf2e6144dd078f2a09c793d
3
+ size 15997
checkpoint-29000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d797d2dc7fe5c96be400c25a4db54eb2f3e8bb330cefce70cf178d25b3ba2fb
3
+ size 15997
checkpoint-29000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c92d2ea071ba5c2943c4da20be681b1298685708bc55b347d89b76c5d303b98
3
+ size 15997
checkpoint-29000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c123e54399317ea68ce3002f4c573e0a38c805f11a64b78ad10619d25a3814a4
3
+ size 15997
checkpoint-29000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fb498d979fc20834825db61c705144a8373bcadeaaf66cb9b8cf8e2f4a8c1e8
3
+ size 15997
checkpoint-29000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2d9f27215c3c1fd95ecf556fae35f426a4678e82c95fddb2d63284bfc970497
3
+ size 15984
checkpoint-29000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1c6ea1ab10d2cc9190d1c373a2a39f129271e93bc4b7eb5a573e14349ba90a5
3
+ size 15984
checkpoint-29000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc5f122136aa4bf0dadc162fa50465c65baed985edbdcc2f97302dac9effed1e
3
+ size 15984
checkpoint-29000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c603b32e81de3bd340d47b65e9679c4372684376a445ed1e898d556f7333ce29
3
+ size 15984
checkpoint-29000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a9fe8949b151f06768bdd24e39679610339d4cc663a9f9ef6a46ae64f7f6120
3
+ size 15984
checkpoint-29000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba5ec92d64393d3f95d617f9ba5c2a9c3dab4dbefe977d28dd7529969266334a
3
+ size 15984
checkpoint-29000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7492a67fc0a3ce492de84ee76df68bc050e990579c4d82817631a7be32045511
3
+ size 15984
checkpoint-29000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62fae8e2769054ebd40bd2b81568acde3805d65ffe388c0842491451fa596d1c
3
+ size 15984
checkpoint-29000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17963727d317dfb508460ed1089817ef3f12b052c232b268d68775fd37f294c6
3
+ size 1064
checkpoint-29000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-29000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e948f3490e211436dd6f621d749b82dca60e77aad37fb5a1079ccb2d7721f35
3
+ size 6008
checkpoint-30000/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": false,
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-30000/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-30000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab51478087a8bff82005b34f04f96b3f50ec29ec9e1e141f758d84fa40b577b0
3
+ size 4538452374
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:059db40341f6b0c2b74f24189c144ed80b21f180ff399f4cbe58c2d82eee4d5f
3
+ size 6630551906
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b625b5221f22537602cbe2b5b25306f0fed9c83fb999ab1779ecd145258db699
3
+ size 15984
checkpoint-30000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96096ba8e65041eae6f1320a4768794a119bf2a35f97010719ed5cf7f64931f0
3
+ size 15984
checkpoint-30000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:979a12e19e90dd363ea83307ebc2baf85fc28b0ba68449159314c47584542823
3
+ size 15997
checkpoint-30000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e8881e7a258100c72b93fe16c82af4c5de3300f5bec24d94176025415a4a8d8
3
+ size 15997
checkpoint-30000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e60dd89a704ef0967cc71dc8ef9291bde5cc3b41b4e65d19ecb3b7a2b1ef012b
3
+ size 15997
checkpoint-30000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f3b0d54f618b661f7ae9408a33f12a081cdfb14af7f98cb91a6f4865f995d9e
3
+ size 15997
checkpoint-30000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f71d153af7fe265b7968eae32166edeb53ae24b93487397fa9922e26ec03d8f
3
+ size 15997
checkpoint-30000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a29e0ac5e63bcb74eaaa41c4357117c3252d6dbfa8aace7a8410777764a68761
3
+ size 15997
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:310a8846a478b0cad83b0bf8c554aa6d57da3893e74ba0a11fe72d8e20fc907d
3
+ size 15984
checkpoint-30000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edcc1fa64515b99cbdf42c4ef35834af761ce4a3feb19bb9ee927dcf94da1818
3
+ size 15984
checkpoint-30000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b38285b6bc2e6ee528a463ef1a8ebf5ee4a4bc113e66410649bd814a46aa668
3
+ size 15984
checkpoint-30000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28b7c51664395177d79fa3eae2542d28d45ed5b4b96a6405672661cc3b7f606a
3
+ size 15984
checkpoint-30000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59579b66b1409a8f2d27c753ac484f7093adfbffa0cc810c64b26890ecbf424e
3
+ size 15984
checkpoint-30000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d56fb2387ec4ed535dcdf59cc07fa3d5c9c68a2541968e1fed66b3afdefdbb3e
3
+ size 15984
checkpoint-30000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:477f1bbaa4d84454a511b6932583b20910ad64e6a134504421d95d694707a1cd
3
+ size 15984
checkpoint-30000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb647d10ccaa4c555615989fc04c188d9e39c9ac2eeb9456ed4a539857356ea6
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10251bcfbd6cb1085d5e9df908000ba794cb50dfd74fc6b90c9196b38832fa8f
3
+ size 1064
checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e948f3490e211436dd6f621d749b82dca60e77aad37fb5a1079ccb2d7721f35
3
+ size 6008