Upload final_model/training_args.json with huggingface_hub
Browse files
final_model/training_args.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"train_file": "dataset/public_phrase/context_filtered/filtered8-vihallu-train.csv",
|
| 3 |
+
"test_file": "dataset/public_phrase/context_filtered/filtered8-vihallu-public-test.csv",
|
| 4 |
+
"output_dir": "./my_results",
|
| 5 |
+
"context_model": "keepitreal/vietnamese-sbert",
|
| 6 |
+
"prompt_model": "keepitreal/vietnamese-sbert",
|
| 7 |
+
"response_model": "keepitreal/vietnamese-sbert",
|
| 8 |
+
"pooling_method": "mean",
|
| 9 |
+
"use_response_prompt_attention": true,
|
| 10 |
+
"chunk_size": 180,
|
| 11 |
+
"chunk_stride": 90,
|
| 12 |
+
"max_chunks": 5,
|
| 13 |
+
"chunk_aggregation": "attention",
|
| 14 |
+
"max_length": 256,
|
| 15 |
+
"batch_size": 16,
|
| 16 |
+
"eval_batch_size": 8,
|
| 17 |
+
"epochs": 12,
|
| 18 |
+
"lr": 1e-05,
|
| 19 |
+
"warmup_steps": 500,
|
| 20 |
+
"weight_decay": 0.01,
|
| 21 |
+
"dropout": 0.3,
|
| 22 |
+
"num_heads": 8,
|
| 23 |
+
"grad_accum_steps": 1,
|
| 24 |
+
"use_rdrop": false,
|
| 25 |
+
"rdrop_alpha": 5.0,
|
| 26 |
+
"eval_steps": 500,
|
| 27 |
+
"save_steps": 500,
|
| 28 |
+
"logging_steps": 1000,
|
| 29 |
+
"patience": 4,
|
| 30 |
+
"seed": 42,
|
| 31 |
+
"use_sbert": true,
|
| 32 |
+
"validation_split": 0.2,
|
| 33 |
+
"sweep": false
|
| 34 |
+
}
|