Actions: THUDM/slime
Actions
Showing runs from all workflows
2,500+ workflow runs
2,500+ workflow runs
_get_capped_partitions crashes when a single sample exceeds max_tokens_per_gpu
Slash Command Handler
#579:
Issue comment #1839 (comment)
created
by
Chios-C
_get_capped_partitions produces empty partitions when num_microbatches is all-reduced across DP ranks
Slash Command Handler
#578:
Issue comment #1838 (comment)
created
by
nameissodifficult
_get_capped_partitions produces empty partitions when num_microbatches is all-reduced across DP ranks
Slash Command Handler
#576:
Issue comment #1838 (comment)
created
by
samaritan1998