diff --git a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_beam.sh b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_beam.sh index db5daa5f2f..b472e30939 100755 --- a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_beam.sh +++ b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_beam.sh @@ -26,7 +26,7 @@ export FUSED_SILU_W13=${FUSED_SILU_W13:-1} export SPLIT_W13=${SPLIT_W13:-0} export OFFLOAD_OPTIM=${OFFLOAD_OPTIM:-0} -export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16" +export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="float32" export DP=${DP:-8} MP=${MP:-1} BS=${BS:-16} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-2} export GBS=$((BS * GRADIENT_ACC_STEPS)) diff --git a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_run.sh b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_run.sh index 0bfe3ae827..bb53cc0ded 100755 --- a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_run.sh +++ b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama31_8b/implementations/tinybox_8xMI350X/dev_run.sh @@ -26,7 +26,7 @@ export FUSED_SILU_W13=${FUSED_SILU_W13:-1} export SPLIT_W13=${SPLIT_W13:-0} export OFFLOAD_OPTIM=${OFFLOAD_OPTIM:-0} -export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16" +export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="float32" export DP=${DP:-8} MP=${MP:-1} BS=${BS:-16} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-2} export GBS=$((BS * GRADIENT_ACC_STEPS))