Update the batch-size reduction factor in find executable batch size (in src/accelerate/utils/memory.py ).
Update the batch-size reduction factor in find executable batch size (in src/accelerate/utils/memory.py ).: a task in HF ML Tasksmith (Harbor dataset). Currently, when an OOM-related exception is caught, the batch size is halved. Change the default retry behavior so that the next batch size is 90%…
Part of FineEnvs/HF_ML_Tasksmith.