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
97a9617
·
verified ·
1 Parent(s): 977d1f1

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-56000/config.json +79 -0
  2. checkpoint-56000/generation_config.json +6 -0
  3. checkpoint-56000/model.safetensors +3 -0
  4. checkpoint-56000/optimizer.pt +3 -0
  5. checkpoint-56000/rng_state_0.pth +3 -0
  6. checkpoint-56000/rng_state_1.pth +3 -0
  7. checkpoint-56000/rng_state_10.pth +3 -0
  8. checkpoint-56000/rng_state_11.pth +3 -0
  9. checkpoint-56000/rng_state_12.pth +3 -0
  10. checkpoint-56000/rng_state_13.pth +3 -0
  11. checkpoint-56000/rng_state_14.pth +3 -0
  12. checkpoint-56000/rng_state_15.pth +3 -0
  13. checkpoint-56000/rng_state_2.pth +3 -0
  14. checkpoint-56000/rng_state_3.pth +3 -0
  15. checkpoint-56000/rng_state_4.pth +3 -0
  16. checkpoint-56000/rng_state_5.pth +3 -0
  17. checkpoint-56000/rng_state_6.pth +3 -0
  18. checkpoint-56000/rng_state_7.pth +3 -0
  19. checkpoint-56000/rng_state_8.pth +3 -0
  20. checkpoint-56000/rng_state_9.pth +3 -0
  21. checkpoint-56000/scheduler.pt +3 -0
  22. checkpoint-56000/trainer_state.json +0 -0
  23. checkpoint-56000/training_args.bin +3 -0
  24. checkpoint-57000/config.json +79 -0
  25. checkpoint-57000/generation_config.json +6 -0
  26. checkpoint-57000/model.safetensors +3 -0
  27. checkpoint-57000/optimizer.pt +3 -0
  28. checkpoint-57000/rng_state_0.pth +3 -0
  29. checkpoint-57000/rng_state_1.pth +3 -0
  30. checkpoint-57000/rng_state_10.pth +3 -0
  31. checkpoint-57000/rng_state_11.pth +3 -0
  32. checkpoint-57000/rng_state_12.pth +3 -0
  33. checkpoint-57000/rng_state_13.pth +3 -0
  34. checkpoint-57000/rng_state_14.pth +3 -0
  35. checkpoint-57000/rng_state_15.pth +3 -0
  36. checkpoint-57000/rng_state_2.pth +3 -0
  37. checkpoint-57000/rng_state_3.pth +3 -0
  38. checkpoint-57000/rng_state_4.pth +3 -0
  39. checkpoint-57000/rng_state_5.pth +3 -0
  40. checkpoint-57000/rng_state_6.pth +3 -0
  41. checkpoint-57000/rng_state_7.pth +3 -0
  42. checkpoint-57000/rng_state_8.pth +3 -0
  43. checkpoint-57000/rng_state_9.pth +3 -0
  44. checkpoint-57000/scheduler.pt +3 -0
  45. checkpoint-57000/trainer_state.json +0 -0
  46. checkpoint-57000/training_args.bin +3 -0
checkpoint-56000/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-56000/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-56000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75c797a7d5abd095a0dbb4ae2385545d863f7a4ca4b69ec2d4428cd72b86b9f6
3
+ size 4538452374
checkpoint-56000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf0d36aeeee4ccd732cd62730eb66a4ca099b6f7d610e15c8adeb0b678f98a86
3
+ size 6630551906
checkpoint-56000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85359baae113ff42e845936fd9f2dc100e1057d5df0c5e88c9115a025c15a269
3
+ size 15984
checkpoint-56000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bffa92d7871aa28744546004b0acd195f4510d3b70af92f0fccacdac82c77d1
3
+ size 15984
checkpoint-56000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56a9e38f37be449253a563bec93d5365be2c7d6b2f07544e06d0cb30c7d860bd
3
+ size 15997
checkpoint-56000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09dbc7373661ee189579342ffdae0b1b491d99b94cdb5696138e08532988b2e3
3
+ size 15997
checkpoint-56000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04a63a0cd36f21df3e611dc40d23ca69f6bf1e322235834faf3ef69c5665641f
3
+ size 15997
checkpoint-56000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:60a0cb4e83897ef1afa0d4dbf4167beaa210de3e086afedb6149a8c7499a67bc
3
+ size 15997
checkpoint-56000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c177dc45c6c42563031fedf7ad34f245cc12f9df9e083627d3a562054d13c25c
3
+ size 15997
checkpoint-56000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:254ebb88b34bd05d51a64e6a7b17cb72143960c8a1dd2820c645d725d6de1d36
3
+ size 15997
checkpoint-56000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7cff5da21bea0ae7ac3a7bdeab9b45dbf02f18132a7c5ea7a4c53d267b3c162
3
+ size 15984
checkpoint-56000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f365fdba5e820a832d5e6c4c9d812e1c91e5e52d388b8412b80cef90a9e7fd4f
3
+ size 15984
checkpoint-56000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6556b58a50200258611cab0e5148d79f6f98f76fbbd46732da1aed25aceaf622
3
+ size 15984
checkpoint-56000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab736a3edcd0dbb836f8cdcef56310e989cd4104b440f1eb11a63753851f95b1
3
+ size 15984
checkpoint-56000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:508ba2fa130dd946b035a4a45eec797dc605bea306ca708feacd28ff802f9955
3
+ size 15984
checkpoint-56000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e673ae9e4806c47549d0a92bb94799a9edec16da1e00b1baae7374d01d50634
3
+ size 15984
checkpoint-56000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2875bfe83638dd7bd9af2a1e4a14702f691c469c6c2c8ebef9f5b17b01c6433
3
+ size 15984
checkpoint-56000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e00d15ab1be34441d48a65a0ec02dd39cd282739eb3326c55c36d0e84cc65c2e
3
+ size 15984
checkpoint-56000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:359327a66d676f34d3152e5056c4db43904ec106736630e54e424fd26f1b7c4c
3
+ size 1064
checkpoint-56000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-56000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01617dfbc1bd4f2397a8f814ab4d1bc1d26b3358a6118dabb8f02a39678cdb57
3
+ size 6008
checkpoint-57000/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-57000/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-57000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96732e2ec8eef29ea6d9eb02938a1b09894be80c0b94c471ebb7cf4f2f8650cd
3
+ size 4538452374
checkpoint-57000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dba5c373f743c81ec5722fa81ddd9b04d764cd7f470cfbf5ca11e6c26acc3fd
3
+ size 6630551906
checkpoint-57000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c80e2d4b4abf594249f5a06e02eb2b69834d8e519656a804d034b1007386b553
3
+ size 15984
checkpoint-57000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8ef2d5a2cebf373dc8d85fddbb2e07cd92e54a8459fefc76bfe5c8c2d394bbb
3
+ size 15984
checkpoint-57000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d43fa0eb1b20eca18b4903c0b5b4ec5f7686c37d78c1faeae8302e790387670e
3
+ size 15997
checkpoint-57000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2620e122bb1bda71a307ee470f5566871eee255b444acec2e36bfe87e65db38e
3
+ size 15997
checkpoint-57000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:118b86c2a4bdab727007cbc7f732ad3d095f270d8a2b8ee91ed4406d3968cbe3
3
+ size 15997
checkpoint-57000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79e00c16d3921aea668de92606d11debb052a7fb56d5043944b1ca9d54b21b8c
3
+ size 15997
checkpoint-57000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2831089f9229d7f84f55cb8cce0e81e2bb652fbcaf6a9ce4fdf3e993efce79a3
3
+ size 15997
checkpoint-57000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c660642c425ca0bc3aa25d5bc23a9d3930a4b10b5246d674efafffa3441690f
3
+ size 15997
checkpoint-57000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d3b2de0aef31c9f74649f335b901c536d2b611e0d712234c273596aa6abc133
3
+ size 15984
checkpoint-57000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:280aa0cf78faa7181e7d768a8998838994f3b1aabb557399f8dd91c3289ce7eb
3
+ size 15984
checkpoint-57000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0640b37fa5ad52cd9a05392fb5253fbe7812a24908189157f738c1552f00e8d3
3
+ size 15984
checkpoint-57000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5113081b53fea4fc06b89edc317f13336785e0e3ae1a47dabdbfeac8ce55d30a
3
+ size 15984
checkpoint-57000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46195c3b2a5d7f79f8c3adf585d2a2fbe1a8df174b6783268430ef1fa6baf95e
3
+ size 15984
checkpoint-57000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:83abe92f5000415fbb19066333539d96abacdc5f4d53b398dc584d5259871b6b
3
+ size 15984
checkpoint-57000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bace84a84c0d1b84f6d1cb50787281339bbb4ccb3a0e757eb548b7184a2eece
3
+ size 15984
checkpoint-57000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9afec8b20ed0e1c55f7a588b9d7d7507bac9a547bb65dcb53939b5e0c1f9d7e1
3
+ size 15984
checkpoint-57000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69b50c8876d7ffe4f62e748723883c79b05f5dc7b526e63515b1252670374a9d
3
+ size 1064
checkpoint-57000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-57000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:005dde0d8ec62655361f895242bfd3b1b52c85dd9385857a32a59f8eb9445220
3
+ size 5944