Activity
another change mbsz back so it is correct for grpo
another change mbsz back so it is correct for grpo
change mbsz back so it is correct for grpo
change mbsz back so it is correct for grpo
fix another vlllm reference and increase timeout
fix another vlllm reference and increase timeout
fix tests.e2e.utils import
fix tests.e2e.utils import
fix one more incorrect schema/config path
fix one more incorrect schema/config path
add gradient checkpointing to multigpu e2e ci
add gradient checkpointing to multigpu e2e ci
update gemma3 model namespace to use mirror
update gemma3 model namespace to use mirror
chore: update pre-commit hooks
chore: update pre-commit hooks
constrain amount of text generated
constrain amount of text generated
chore: refactor use_cache handling
chore: refactor use_cache handling
configurable heads_k_stride from ring-flash-attn hf adapter
configurable heads_k_stride from ring-flash-attn hf adapter
Force push
fix set/unset of hub var
fix set/unset of hub var