We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent d7613f4 commit a239f0cCopy full SHA for a239f0c
src/forge/actors/trainer.py
@@ -176,8 +176,7 @@ async def train_step(
176
177
# TODO: delete item() to avoid cpu-gpu sync
178
loss = loss.detach().item()
179
- record_metric("rl_trainer/count_training_steps", 1, Reduce.SUM)
180
- record_metric("rl_trainer/avg_grpo_loss", loss, Reduce.MEAN)
+ record_metric("rl_trainer/avg_loss", loss, Reduce.MEAN)
181
182
# These are placeholder values until the loss function exposes these metrics
183
# record_metric("rl_trainer/step/avg_kl_divergence", 0.0, Reduce.MEAN)
0 commit comments