tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

update HF HUB env var and fix reward trainer log since it doesn't directly override log

Browse Source
This commit is contained in:
Wing Lian
2024-12-06 14:52:59 -05:00
parent 3b44989205
commit 825f66b9fd
3 changed files with 2 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
docker/Dockerfile-cloud
View File

@@ -2,7 +2,7 @@ ARG BASE_TAG=main
FROM axolotlai/axolotl:$BASE_TAG
ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
ENV HF_HUB_CACHE="/workspace/data/huggingface-cache/hub"
ENV HF_HOME="/workspace/data/huggingface-cache/hub"
ENV HF_HUB_ENABLE_HF_TRANSFER="1"

2
docker/Dockerfile-cloud-no-tmux
View File

@@ -2,7 +2,7 @@ ARG BASE_TAG=main
FROM axolotlai/axolotl:$BASE_TAG
ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
ENV HF_HUB_CACHE="/workspace/data/huggingface-cache/hub"
ENV HF_HOME="/workspace/data/huggingface-cache/hub"
ENV HF_HUB_ENABLE_HF_TRANSFER="1"

6
src/axolotl/core/trainer_builder.py
View File

@@ -1266,12 +1266,6 @@ class AxolotlRewardTrainer(SchedulerMixin, RewardTrainer):
def log(self, logs: Dict[str, float], start_time: Optional[float] = None) -> None:
# TODO remove once trl supports the updated to the Trainer.log method
# logs either has 'loss' or 'eval_loss'
train_eval = "train" if "loss" in logs else "eval"
# Add averaged stored metrics to logs
for key, metrics in self._stored_metrics[train_eval].items():
logs[key] = torch.tensor(metrics).mean().item()
del self._stored_metrics[train_eval]
return super(RewardTrainer, self).log( # pylint: disable=bad-super-call
logs, start_time
)