부창규

I have read the CLA Document and I sign the CLA

I did a mistake.. I will reopen pr

Thank you for your kind response. I already resubmit pr but I will read contribution guide. It's not that easy for beginner who are not familiar to such process.

Hello. 1) In agentlightning/verl/trainer.py 346th line, ```python if self.config.trainer.balance_batch: self._balance_batch(batch, metrics=metrics) # update critic if self.use_critic: with _timer("update_critic", timing_raw): -> critic_output = self.critic_wg.update_critic(batch) critic_output_metrics = reduce_metrics(critic_output.meta_info["metrics"]) metrics.update(critic_output_metrics) ``` below does...

Thanks I misunderstood that batch is not affected. I should read balance_batch func and trainer parts carefully.

In my case, This helped. ```python curl -LsSf https://astral.sh/uv/install.sh | sh export PATH="$HOME/.local/bin:$PATH" ```

> @genji970 i think the issue is "sys_platform == 'linux'" line Yes. Just for the first condition, it worked in my case. So just mentioned.

model.val(), benchmark make error benchmark :   | Format | Status❔ | Size (MB) | metrics/mAP50-95(B) | Inference time (ms/im) | FPS -- | -- | -- | -- | --...

Do you mean that if I do model = YOLO(yaml or model path) model.predict() then model automatically estimate running time for embed? I'm a bit confused that when I code...

When I look at utils.benchmark.py, it seems like it does not get embed for inputs. Do you mean that I can change predict function sourcecode(embed : None -> embed:[1,2,3]) and...