Upload trainer.json with huggingface_hub
Browse files- trainer.json +30 -0
trainer.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"repo_id": "habanoz/wiki-text-gpt-v0.11",
|
| 3 |
+
"trainer_config": {
|
| 4 |
+
"seq_length": 512,
|
| 5 |
+
"gradient_accumulation_steps": 1,
|
| 6 |
+
"batch_size": 32,
|
| 7 |
+
"data_dir": "eco-news-tr",
|
| 8 |
+
"warmup_iters": 100,
|
| 9 |
+
"learning_rate": 0.001,
|
| 10 |
+
"lr_decay_iters": 5000,
|
| 11 |
+
"max_iters": 5000,
|
| 12 |
+
"min_lr": 0.0001,
|
| 13 |
+
"weight_decay": 0.1,
|
| 14 |
+
"beta1": 0.9,
|
| 15 |
+
"beta2": 0.99,
|
| 16 |
+
"compile": false,
|
| 17 |
+
"decay_lr": true,
|
| 18 |
+
"seed": 147,
|
| 19 |
+
"log_interval": 10,
|
| 20 |
+
"eval_interval": 250,
|
| 21 |
+
"eval_iters": 200,
|
| 22 |
+
"out_dir": "wiki-text-gpt-v0.11",
|
| 23 |
+
"wandb_log": true,
|
| 24 |
+
"wandb_project": "NB-Haber-GPT-Training",
|
| 25 |
+
"wandb_run_name": "haber-gpt-v1.021(wikitr-tokenizer-seed-147)",
|
| 26 |
+
"wandb_run_id": "1721342972",
|
| 27 |
+
"grad_norm_clip": 1.0,
|
| 28 |
+
"dtype": "float16"
|
| 29 |
+
}
|
| 30 |
+
}
|