58 lines
1.5 KiB
JSON
58 lines
1.5 KiB
JSON
|
{
|
||
|
|
"student_model": "unsloth/Qwen2.5-3B-Instruct-bnb-4bit",
|
||
|
|
"teacher_model": "Qwen/Qwen2.5-14B-Instruct",
|
||
|
|
"teacher_fallback": "Qwen/Qwen2.5-7B-Instruct",
|
||
|
|
"teacher_data_parallel": "True",
|
||
|
|
"use_agnews": "True",
|
||
|
|
"n_agnews": "300",
|
||
|
|
"use_snli": "True",
|
||
|
|
"n_snli": "200",
|
||
|
|
"use_mnli": "True",
|
||
|
|
"n_mnli": "150",
|
||
|
|
"use_anli": "True",
|
||
|
|
"n_anli": "200",
|
||
|
|
"use_fever": "True",
|
||
|
|
"n_fever": "150",
|
||
|
|
"use_liar": "True",
|
||
|
|
"n_liar": "150",
|
||
|
|
"n_synth": "600",
|
||
|
|
"gen_batch": "8",
|
||
|
|
"gen_max_new": "380",
|
||
|
|
"label_batch": "8",
|
||
|
|
"label_max_new": "900",
|
||
|
|
"max_chars": "3500",
|
||
|
|
"max_seq_length": "3072",
|
||
|
|
"lora_r": "32",
|
||
|
|
"lora_alpha": "32",
|
||
|
|
"epochs": "3",
|
||
|
|
"batch_size": "2",
|
||
|
|
"grad_accum": "8",
|
||
|
|
"lr": "0.0002",
|
||
|
|
"warmup": "10",
|
||
|
|
"save_steps": "50",
|
||
|
|
"save_total_limit": "2",
|
||
|
|
"eval_steps": "25",
|
||
|
|
"load_best": "True",
|
||
|
|
"balance_cap": "400",
|
||
|
|
"gold_upweight": "3",
|
||
|
|
"val_fraction": "0.08",
|
||
|
|
"seed": "3407",
|
||
|
|
"resume": "True",
|
||
|
|
"min_keep_rate": "0.4",
|
||
|
|
"min_valid_examples": "400",
|
||
|
|
"gate_heldout_validity": "0.7",
|
||
|
|
"gate_unseen_validity": "0.5",
|
||
|
|
"min_avg_output_chars": "60",
|
||
|
|
"retry_attempts": "3",
|
||
|
|
"retry_delay": "5",
|
||
|
|
"min_disk_free_gb": "12.0",
|
||
|
|
"rows_ram_cap": "4000",
|
||
|
|
"cleanup_after_push": "True",
|
||
|
|
"hf_push": "True",
|
||
|
|
"hf_repo_id": "Luimas/claim-extractor-detective-qwen3b",
|
||
|
|
"hf_private": "False",
|
||
|
|
"hf_data_repo_id": "Luimas/claim-extractor-gold-corpus",
|
||
|
|
"hf_data_push": "True",
|
||
|
|
"bundle_release_zip": "True",
|
||
|
|
"base_dir": "/kaggle/working/claim_extractor_v5"
|
||
|
|
}
|