Upload decoder/small_32gpus/decoder_config.json with huggingface_hub
Browse files
decoder/small_32gpus/decoder_config.json
CHANGED
|
@@ -16,7 +16,7 @@
|
|
| 16 |
"attn_dim_head": 64,
|
| 17 |
"sparse_attn": true,
|
| 18 |
"memory_efficient": true,
|
| 19 |
-
|
| 20 |
}
|
| 21 |
],
|
| 22 |
"clip": {
|
|
@@ -57,13 +57,13 @@
|
|
| 57 |
},
|
| 58 |
"train": {
|
| 59 |
"epochs": 1000,
|
| 60 |
-
"lr":
|
| 61 |
"wd": 0.01,
|
| 62 |
"max_grad_norm": 0.5,
|
| 63 |
"save_every_n_samples": 2000000,
|
| 64 |
"n_sample_images": 10,
|
| 65 |
"device": "cuda:0",
|
| 66 |
-
"epoch_samples":
|
| 67 |
"validation_samples": 100000,
|
| 68 |
"use_ema": true,
|
| 69 |
"ema_beta": 0.99,
|
|
@@ -91,15 +91,16 @@
|
|
| 91 |
|
| 92 |
"wandb_entity": "rom1504",
|
| 93 |
"wandb_project": "dalle2_train_decoder",
|
| 94 |
-
"wandb_resume":
|
|
|
|
| 95 |
|
| 96 |
"verbose": true
|
| 97 |
},
|
| 98 |
|
| 99 |
"load": {
|
| 100 |
-
|
| 101 |
"load_from": "url",
|
| 102 |
-
"url": "https://api.wandb.ai/files/rom1504/dalle2_train_decoder/
|
| 103 |
},
|
| 104 |
|
| 105 |
"save": [{
|
|
@@ -108,11 +109,12 @@
|
|
| 108 |
{
|
| 109 |
"save_to": "huggingface",
|
| 110 |
"huggingface_repo": "laion/DALLE2-PyTorch",
|
| 111 |
-
|
| 112 |
|
| 113 |
"save_latest_to": "decoder/small_32gpus/latest.pth",
|
| 114 |
|
| 115 |
"save_type": "model"
|
| 116 |
}]
|
| 117 |
}
|
| 118 |
-
}
|
|
|
|
|
|
| 16 |
"attn_dim_head": 64,
|
| 17 |
"sparse_attn": true,
|
| 18 |
"memory_efficient": true,
|
| 19 |
+
"self_attn": [false, true, true, true]
|
| 20 |
}
|
| 21 |
],
|
| 22 |
"clip": {
|
|
|
|
| 57 |
},
|
| 58 |
"train": {
|
| 59 |
"epochs": 1000,
|
| 60 |
+
"lr": 5e-5,
|
| 61 |
"wd": 0.01,
|
| 62 |
"max_grad_norm": 0.5,
|
| 63 |
"save_every_n_samples": 2000000,
|
| 64 |
"n_sample_images": 10,
|
| 65 |
"device": "cuda:0",
|
| 66 |
+
"epoch_samples": 8000000,
|
| 67 |
"validation_samples": 100000,
|
| 68 |
"use_ema": true,
|
| 69 |
"ema_beta": 0.99,
|
|
|
|
| 91 |
|
| 92 |
"wandb_entity": "rom1504",
|
| 93 |
"wandb_project": "dalle2_train_decoder",
|
| 94 |
+
"wandb_resume": true,
|
| 95 |
+
"wandb_run_id": "2yi3dyra",
|
| 96 |
|
| 97 |
"verbose": true
|
| 98 |
},
|
| 99 |
|
| 100 |
"load": {
|
| 101 |
+
"resume": true,
|
| 102 |
"load_from": "url",
|
| 103 |
+
"url": "https://api.wandb.ai/files/rom1504/dalle2_train_decoder/2yi3dyra/latest.pth"
|
| 104 |
},
|
| 105 |
|
| 106 |
"save": [{
|
|
|
|
| 109 |
{
|
| 110 |
"save_to": "huggingface",
|
| 111 |
"huggingface_repo": "laion/DALLE2-PyTorch",
|
| 112 |
+
"save_meta_to": "decoder/small_32gpus/",
|
| 113 |
|
| 114 |
"save_latest_to": "decoder/small_32gpus/latest.pth",
|
| 115 |
|
| 116 |
"save_type": "model"
|
| 117 |
}]
|
| 118 |
}
|
| 119 |
+
}
|
| 120 |
+
|