Search results

请问1.5B的训练过程format reward一直是0吗

Status: Open.

#73 In dhcode-cpp/X-R1;

· zhaoyang123638 opened on Mar 18, 2026

The hostname of the client socket cannot be retrieved. err=-3

Status: Open.

#71 In dhcode-cpp/X-R1;

· Minami-su opened on Jul 11, 2025

瓜神您好，有关num_generations设置以及代码运行时会hang住的问题

Status: Open.

#70 In dhcode-cpp/X-R1;

· ZhangEnmao opened on Apr 18, 2025

/usr/bin/ld: cannot find -lcurand: No such file or directory

Status: Open.

#68 In dhcode-cpp/X-R1;

· WANGDUNDUN1 opened on Mar 26, 2025

Using lora for training is not effective

Status: Open.

#66 In dhcode-cpp/X-R1;

· tsrigo opened on Mar 18, 2025

RuntimeError: 'weight' must be 2-D

Status: Open.

#65 In dhcode-cpp/X-R1;

· hengzhi40 opened on Mar 11, 2025

So many bugs in the new version code

Status: Open.

#64 In dhcode-cpp/X-R1;

· yangluo23 opened on Mar 10, 2025

ValueError: You are attempting to perform batched generation with padding_side='right' this may lead to unexpected behaviour for Flash Attention version of Qwen2. Make sure to call tokenizer.padding_side = 'left' before tokenizing the input.

Status: Open.

#63 In dhcode-cpp/X-R1;

· nioxinjiang3 opened on Mar 10, 2025

SCQ5K，训练出来的曲线看不懂啊？

Status: Open.

#62 In dhcode-cpp/X-R1;

· gumanchang opened on Mar 9, 2025

self.accelerator.unwrap_model(self.model).disable_adapters() ??

Status: Open.

#61 In dhcode-cpp/X-R1;

· mysuns opened on Mar 2, 2025

Use Qwen2.5-3B-Instruct get warning: flash attention 2.0 with a model not initialized on GPU

Status: Open.

#60 In dhcode-cpp/X-R1;

· ShawFeng99 opened on Feb 26, 2025

训练过程中Qwen-7B模型输出无法获得有效结果，使得Reward始终为0.

Status: Open.

#59 In dhcode-cpp/X-R1;

· taobiaoli1314 opened on Feb 26, 2025