laion/CoderForge-Preview-v3-100000
收藏Hugging Face2026-04-22 更新2026-04-26 收录
下载链接：
https://hf-mirror.com/datasets/laion/CoderForge-Preview-v3-100000
下载链接
链接失效反馈官方服务：
资源简介：
laion/CoderForge-Preview-v3-100000数据集是togethercomputer/CoderForge-Preview数据集的一个子集，包含100,000行预处理的tokenized数据。数据格式包括input_ids、attention_mask、labels、chat_template_applied、trajectory_id、reward和source等列。数据集主要用于文本生成任务，支持Qwen3的tokenizer。数据是从4个源slug（R2E_Gym、SWE_Rebench、SWE_Smith、filtered_reward1）中确定性采样（seed=42）并连接而成的。
laion/CoderForge-Preview-v3-100000 is a row-subset of the pre-tokenized trajectories in togethercomputer/CoderForge-Preview (trajectories-tokenized_qwencoder subset). It contains 100,000 rows of pre-tokenized data for Qwen3 (tokenizer shared with Qwen2.5-Coder / Qwen3-Coder / Qwen3-8B). Each row includes columns such as input_ids, attention_mask, labels, chat_template_applied, trajectory_id, reward, and source. The data is sampled deterministically (seed=42) from a concatenation of all 4 source slugs (R2E_Gym, SWE_Rebench, SWE_Smith, filtered_reward1).
提供机构：
laion
