!210 修复datasets相关文档

Merge pull request !210 from 幽若/master-docfix-520
This commit is contained in:
2025-05-20 09:29:43 +00:00
committed by i-robot
parent 40823839b7
commit ecfa61c360
3 changed files with 5 additions and 5 deletions

View File

@ -29,7 +29,7 @@ dataset: alpaca_zh_51k
### 数据处理
openMind目前支持Alpaca、ShareGPTText种数据格式,自定义数据集需要转换为这种格式之一。各格式支持的训练阶段如下:
openMind目前支持Alpaca、ShareGPTText和Pairwise四种数据格式,自定义数据集需要转换为这种格式之一。各格式支持的训练阶段如下:
<table>
<thead>
<tr>

View File

@ -32,7 +32,7 @@ dataset: rlhf-reward-datasets
cutoff_len: 1024
# output
output_dir: saves/qwen2_7b_reward
output_dir: saves/qwen2_7b_dpo
logging_steps: 1
save_steps: 20000
overwrite_output_dir: true
@ -63,7 +63,7 @@ from openmind import run_train
run_train(
model_name_or_path = "/mnt/h/pretrain_models/Qwen2.5-0.5B/",
stage="rm",
stage="dpo",
template="qwen",
do_train=True,
finetuning_type="lora",
@ -71,7 +71,7 @@ run_train(
lora_rank=8,
lora_alpha=16,
dataset="rlhf-reward-datasets",
output_dir="saves/qwen2.5_0.5b_lora_rm",
output_dir="saves/qwen2.5_0.5b_lora_dpo",
logging_steps=1,
save_steps=20000,
overwrite_output_dir=True,

View File

@ -8,7 +8,7 @@ openMind Library当前已支持reward训练用户可通过如下步骤启动r
openMind Library命令行接口内置于openMind Library中安装openMind Library即可使用详细步骤参考[openMind Library安装指南](../../../../install.md)。
*`注openMind进行dpo训练依赖trl>=0.16.1datasets >= 2.18.0, <= 2.21.0openMind和trl两者存在datasets版本依赖冲突请在安装完trl后手动安装datasets对应版本。`*
*`注openMind进行reward训练依赖trl>=0.16.1datasets >= 2.18.0, <= 2.21.0openMind和trl两者存在datasets版本依赖冲突请在安装完trl后手动安装datasets对应版本。`*
## 模型微调示例