Add config.json (2025-05-06T20-38-18Z)
Browse files
artifacts/2025-05-06T20-38-18Z/config.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"seed": 42,
|
| 3 |
+
"hf_dataset_id": "tsilva/clinical-field-mappings",
|
| 4 |
+
"hf_source_model_id": "distilbert/distilgpt2",
|
| 5 |
+
"hf_target_model_id": "tsilva/clinical-field-mapper-classification",
|
| 6 |
+
"export_onnx": true,
|
| 7 |
+
"manual_test_on_end": true,
|
| 8 |
+
"pytorch_compile": true,
|
| 9 |
+
"task_type": "classification",
|
| 10 |
+
"freeze_base_model": false,
|
| 11 |
+
"trainer": {
|
| 12 |
+
"per_device_train_batch_size": 1024,
|
| 13 |
+
"per_device_eval_batch_size": 1024,
|
| 14 |
+
"gradient_accumulation_steps": 1,
|
| 15 |
+
"num_train_epochs": 50,
|
| 16 |
+
"learning_rate": 0.0005,
|
| 17 |
+
"lr_scheduler_type": "cosine",
|
| 18 |
+
"warmup_ratio": 0.01,
|
| 19 |
+
"optim": "adamw_bnb_8bit",
|
| 20 |
+
"weight_decay": 0.1,
|
| 21 |
+
"max_grad_norm": 0,
|
| 22 |
+
"fp_datatype": "fp16",
|
| 23 |
+
"mixed_precision_training": true,
|
| 24 |
+
"fp16": true,
|
| 25 |
+
"bf16": false,
|
| 26 |
+
"tf32": false,
|
| 27 |
+
"fp16_full_eval": true,
|
| 28 |
+
"bf16_full_eval": false,
|
| 29 |
+
"enable_early_stopping": true,
|
| 30 |
+
"early_stopping_patience": 5,
|
| 31 |
+
"early_stopping_threshold": 0.0,
|
| 32 |
+
"gradient_checkpointing": false,
|
| 33 |
+
"use_deepspeed": true
|
| 34 |
+
},
|
| 35 |
+
"run_timestamp": "2025-05-06T20-38-18Z"
|
| 36 |
+
}
|