Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-ae-f32c32-1.0-768_patch-1_group-7

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
cba3a7b
·
verified ·
1 Parent(s): e8fa466

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-10000/config.json +79 -0
  2. checkpoint-10000/generation_config.json +6 -0
  3. checkpoint-10000/model.safetensors +3 -0
  4. checkpoint-10000/optimizer.pt +3 -0
  5. checkpoint-10000/rng_state_0.pth +3 -0
  6. checkpoint-10000/rng_state_1.pth +3 -0
  7. checkpoint-10000/rng_state_10.pth +3 -0
  8. checkpoint-10000/rng_state_11.pth +3 -0
  9. checkpoint-10000/rng_state_12.pth +3 -0
  10. checkpoint-10000/rng_state_13.pth +3 -0
  11. checkpoint-10000/rng_state_14.pth +3 -0
  12. checkpoint-10000/rng_state_15.pth +3 -0
  13. checkpoint-10000/rng_state_2.pth +3 -0
  14. checkpoint-10000/rng_state_3.pth +3 -0
  15. checkpoint-10000/rng_state_4.pth +3 -0
  16. checkpoint-10000/rng_state_5.pth +3 -0
  17. checkpoint-10000/rng_state_6.pth +3 -0
  18. checkpoint-10000/rng_state_7.pth +3 -0
  19. checkpoint-10000/rng_state_8.pth +3 -0
  20. checkpoint-10000/rng_state_9.pth +3 -0
  21. checkpoint-10000/scheduler.pt +3 -0
  22. checkpoint-10000/trainer_state.json +0 -0
  23. checkpoint-10000/training_args.bin +3 -0
  24. checkpoint-9000/config.json +79 -0
  25. checkpoint-9000/generation_config.json +6 -0
  26. checkpoint-9000/model.safetensors +3 -0
  27. checkpoint-9000/optimizer.pt +3 -0
  28. checkpoint-9000/rng_state_0.pth +3 -0
  29. checkpoint-9000/rng_state_1.pth +3 -0
  30. checkpoint-9000/rng_state_10.pth +3 -0
  31. checkpoint-9000/rng_state_11.pth +3 -0
  32. checkpoint-9000/rng_state_12.pth +3 -0
  33. checkpoint-9000/rng_state_13.pth +3 -0
  34. checkpoint-9000/rng_state_14.pth +3 -0
  35. checkpoint-9000/rng_state_15.pth +3 -0
  36. checkpoint-9000/rng_state_2.pth +3 -0
  37. checkpoint-9000/rng_state_3.pth +3 -0
  38. checkpoint-9000/rng_state_4.pth +3 -0
  39. checkpoint-9000/rng_state_5.pth +3 -0
  40. checkpoint-9000/rng_state_6.pth +3 -0
  41. checkpoint-9000/rng_state_7.pth +3 -0
  42. checkpoint-9000/rng_state_8.pth +3 -0
  43. checkpoint-9000/rng_state_9.pth +3 -0
  44. checkpoint-9000/scheduler.pt +3 -0
  45. checkpoint-9000/trainer_state.json +0 -0
  46. checkpoint-9000/training_args.bin +3 -0
checkpoint-10000/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": 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-10000/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-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a7c64010ec406cd317cfee0d856a43c96d454b417c02a2717c8822115839b18
3
+ size 4559801712
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09c2d6336e4b95d644dd0ecb3dc45f8b4829a0619bae0cec76b0138ba1b4ed3d
3
+ size 6628766946
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abff8b36d7e6eb1de792dc9602ee1566e567942c811544ab4a506dcf532f2c44
3
+ size 15984
checkpoint-10000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36cb5bce5edf014623e90a9faf208b1713b312056eba493d2992c1d1846c5fa6
3
+ size 15984
checkpoint-10000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:effb968984f27038eef91a53c47563d564f20b9370020d689786d4e614bf53ea
3
+ size 15997
checkpoint-10000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6fd2b941e3e35c5e64fa14527dceaf81a4a3a6c15dae7f8388d7858afcb69905
3
+ size 15997
checkpoint-10000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d80ace9d7d6b507fbd11b8e229968d5bb00e6bda3cb385317759bcfea1836028
3
+ size 15997
checkpoint-10000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4bdbd3a6778454a9f2e2fd2a731d05673a9b6f43fc429c5561d338256b712d6
3
+ size 15997
checkpoint-10000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:325d37b2fd31bf03bbe791aba3ea2603a59270dbc1f037d58f93e4f04ebe7caf
3
+ size 15997
checkpoint-10000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea84028c84a08338d63321ede28eb68aaed4aedae4e7b966471d9946f62a9b05
3
+ size 15997
checkpoint-10000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b39b7f2a9f9503fbe17e18f7e74b3892e4e93c38c19640219396470e1a8087ea
3
+ size 15984
checkpoint-10000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b6e8ec7057b89e7f53d6939a7328812b41833e2f3180dc61681aaa1040abf85
3
+ size 15984
checkpoint-10000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3882bdf18f80a19c6a49e9741269400d97505708aa6f8d9a487fa68d2113b6b9
3
+ size 15984
checkpoint-10000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0745868678e7692c0bc340bf8534f678482dcebe7dd49c1a078b2cb0c0d9474e
3
+ size 15984
checkpoint-10000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da42fde5ad37710bb88cd2a144a01ba59b4484ae51a3bbc9bfb25b8058ba5ae2
3
+ size 15984
checkpoint-10000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:878b594430fc71fbc3b6c3f082a1c215f40ffe79a254b2149076c806505ed6e6
3
+ size 15984
checkpoint-10000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:601dedf3e831ddaeba826a1c7a44d6578ec5535afe109e8c9052ba3088956e8b
3
+ size 15984
checkpoint-10000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6aaf1823e8ff3928d5db66d9c9f3a51e2e207107c44a3e6a3b0263966fabd9b
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32307d7b1f22786cae513d730e9f747e03deb05c8f6da497266876a0adb16662
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd9ffdb582c86fd6d1f89ddeac8c919b4bf1405e3e8547265350494d1eeb702c
3
+ size 6008
checkpoint-9000/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": 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-9000/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-9000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f1b84d5116f9a6b9bee9ab9233948512c4cccb1844603fe464ffc0ba6b93568
3
+ size 4559801712
checkpoint-9000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd838507f50490132090dacf1f2cdbc0cb226ec537e93c31ad9debebf107ed10
3
+ size 6628766946
checkpoint-9000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c06b9d723d71d7ce4cfbf46fa5725f307b34aeaca18229654daaaca8ced56158
3
+ size 15984
checkpoint-9000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bef5467da023236a9a8bcc183141fa22a98fa0797a3a9a37528a1a7889c17624
3
+ size 15984
checkpoint-9000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e47411d0f509b8a3acd68af5a47139b6492e766b85ab62bf56cd6278cba750d
3
+ size 15997
checkpoint-9000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:069600b5ae742d692f49f0108daaf521f6224edb396617a56e8d4846cadc95ba
3
+ size 15997
checkpoint-9000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc150b4a71db2405b5777b4f9dd67a54af426ee9a67f66384e128776f83965cd
3
+ size 15997
checkpoint-9000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:512b899ac7df1f4bc1bab199a0af5b0623166ee76dadceaa4bf00feb1373f6eb
3
+ size 15997
checkpoint-9000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b76eb8d9451d7663e09eca060d0f3b0a6ac0cdb3c57d1ede1e65c6a0332f4a5a
3
+ size 15997
checkpoint-9000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee8ec739fb4dbb8352028bd402e3c45de417026d730549cd7baa9b51e1e188f3
3
+ size 15997
checkpoint-9000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:302365b1f8604d98622660fed4e5bd56de22af9b60b7feb34141bd0e0e221ea1
3
+ size 15984
checkpoint-9000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c196f223ff71dd3512195330d7ab81f673f77a349447715c635e5a83c3ca9eee
3
+ size 15984
checkpoint-9000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b59e3a8aaf3a25e5e1a5dbc66ef1116d088044fee02b0cc0dd68f17bc82e9a2
3
+ size 15984
checkpoint-9000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2d4dd18ecd4ba24acdb22cc358e320132d1ef8536deabe7e02ab78575e96269
3
+ size 15984
checkpoint-9000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08ecd737cbfea91aa1616840e21cf0f917cda114afbd5266507ccc9196f30263
3
+ size 15984
checkpoint-9000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:906f30b8258023b7b8c10fadea3256cffdc0205166578f58b476a7181ff29e36
3
+ size 15984
checkpoint-9000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bad4e25e23d2c0914733bfe2305174bdf02ae1f046360ffa956bd96f26f9ff2b
3
+ size 15984
checkpoint-9000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:015d489b3a041333e0a6f5784769b78c4b7feab5d379d4b54f59f252dda17f0f
3
+ size 15984
checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c44982f2543403d7b017b781199ed567414ec7d61e61e1e7a5e200f318a447b
3
+ size 1064
checkpoint-9000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd9ffdb582c86fd6d1f89ddeac8c919b4bf1405e3e8547265350494d1eeb702c
3
+ size 6008