Upload model_params.json with huggingface_hub
Browse files- model_params.json +30 -0
model_params.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"cache_dir": "/leonardo_work/EUHPC_E03_068/.cache",
|
| 3 |
+
"method": "orpo",
|
| 4 |
+
"dataset": "autoredteam",
|
| 5 |
+
"model": "meta-llama/Llama-3.2-1B",
|
| 6 |
+
"tokenizer": "meta-llama/Llama-3.2-1B-instruct",
|
| 7 |
+
"train_data_path": "/leonardo_work/EUHPC_E03_068/datasets/working/autoredteam_helpfulness_v1-train.jsonl",
|
| 8 |
+
"test_data_path": "/leonardo_work/EUHPC_E03_068/datasets/working/autoredteam_helpfulness_v1-test.jsonl",
|
| 9 |
+
"lr": 0.01,
|
| 10 |
+
"train_batch_size": 300,
|
| 11 |
+
"eval_batch_size": 300,
|
| 12 |
+
"num_epochs": 1,
|
| 13 |
+
"seed": 42,
|
| 14 |
+
"eval_only": false,
|
| 15 |
+
"evaluation_size": null,
|
| 16 |
+
"gradient_accumulation_steps": 2,
|
| 17 |
+
"checkpoint_path": null,
|
| 18 |
+
"experiment_name": "Llama_3.2_1b-autoredteam_helpfulness-train",
|
| 19 |
+
"experiment_group": "results",
|
| 20 |
+
"reference_model": null,
|
| 21 |
+
"context_length": 1024,
|
| 22 |
+
"train_summarization": "",
|
| 23 |
+
"dpo_beta": 0.1,
|
| 24 |
+
"orpo_beta": 0.1,
|
| 25 |
+
"kl_coef": 0.0,
|
| 26 |
+
"reward_model": "",
|
| 27 |
+
"bestofn_size": 4,
|
| 28 |
+
"train_reward_model": "",
|
| 29 |
+
"max_steps": 5511
|
| 30 |
+
}
|