diff --git a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_beam.sh b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_beam.sh
index 22f15f5f87..8885fd58a7 100755
--- a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_beam.sh
+++ b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_beam.sh
@@ -16,6 +16,10 @@ export MASTER_WEIGHTS=${MASTER_WEIGHTS:-1}
 export FP8=${FP8:-1}
 export ALLREDUCE_CAST=${ALLREDUCE_CAST:-1}
 export FAST_CE=${FASE_CE:-1}
+export FUSED_INPUT_QUANTIZE=${FUSED_INPUT_QUANTIZE:-1}
+export FUSED_ADD_NORM_MUL_QUANTIZE=${FUSED_ADD_NORM_MUL_QUANTIZE:-1}
+export FUSED_SILU_W13=${FUSED_SILU_W13:-1}
+export FUSED_PAD_GRAD_ACCUM=${FUSED_PAD_GRAD_ACCUM:-1}
 
 export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
 export DP=${DP:-8} MP=${MP:-1} BS=${BS:-16} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-2}
diff --git a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_run.sh b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_run.sh
index cff289dde5..e68218d2e5 100755
--- a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_run.sh
+++ b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/dev_run.sh
@@ -16,6 +16,10 @@ export MASTER_WEIGHTS=${MASTER_WEIGHTS:-1}
 export FP8=${FP8:-1}
 export ALLREDUCE_CAST=${ALLREDUCE_CAST:-1}
 export FAST_CE=${FASE_CE:-1}
+export FUSED_INPUT_QUANTIZE=${FUSED_INPUT_QUANTIZE:-1}
+export FUSED_ADD_NORM_MUL_QUANTIZE=${FUSED_ADD_NORM_MUL_QUANTIZE:-1}
+export FUSED_SILU_W13=${FUSED_SILU_W13:-1}
+export FUSED_PAD_GRAD_ACCUM=${FUSED_PAD_GRAD_ACCUM:-1}
 
 export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
 export DP=${DP:-8} MP=${MP:-1} BS=${BS:-16} EVAL_BS=${EVAL_BS:-8} GRADIENT_ACC_STEPS=${GRADIENT_ACC_STEPS:-2}
diff --git a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/run_and_time.sh b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/run_and_time.sh
index 0851d92928..fbff01a222 100755
--- a/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/run_and_time.sh
+++ b/examples/mlperf/training_submission_v6.0/tinycorp/benchmarks/llama8b/implementations/tinybox_8xMI350X/run_and_time.sh
@@ -17,6 +17,10 @@ export MASTER_WEIGHTS=1
 export FP8=1
 export ALLREDUCE_CAST=1
 export FAST_CE=1
+export FUSED_INPUT_QUANTIZE=1
+export FUSED_ADD_NORM_MUL_QUANTIZE=1
+export FUSED_SILU_W13=1
+export FUSED_PAD_GRAD_ACCUM=1
 
 export DEFAULT_FLOAT="bfloat16" OPTIM_DTYPE="bfloat16"
 export DP=8 MP=1 BS=16 EVAL_BS=8 GRADIENT_ACC_STEPS=2