初始化项目,由ModelHub XC社区提供模型

Model: QpiEImitation/gkd_gsm8k_S-Qwen2.5-3B-Instruct_T-Qwen2-7B-Instruct
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-06-16 04:34:18 +08:00
commit 35e809098a
10 changed files with 299 additions and 0 deletions

20
run_config.json Normal file
View File

@@ -0,0 +1,20 @@
{
"method": "gkd",
"student_model": "Qwen/Qwen2.5-3B-Instruct",
"teacher_model": "Qwen/Qwen2-7B-Instruct",
"dataset": "gsm8k/main",
"train_samples": 2000,
"eval_samples": 256,
"beta": 0.5,
"max_steps": 800,
"num_epochs": 1.0,
"lr": 2e-06,
"batch_size": 1,
"grad_accum": 8,
"output_dir": "gkd_gsm8k_S-Qwen2.5-3B-Instruct_T-Qwen2-7B-Instruct_20260420_154428",
"use_lora": false,
"lora_r": null,
"lora_alpha": null,
"lora_dropout": null,
"lora_target_modules": null
}