@@ -62,6 +62,11 @@ for qformat in $QFORMAT; do
6262done
6363IFS=" "
6464
65+ if [ -n " $RECIPE " ] && [ -n " $QFORMAT " ]; then
66+ echo " Error: --recipe and --quant are mutually exclusive." >&2
67+ exit 1
68+ fi
69+
6570script_dir=" $( dirname " $( readlink -f " $0 " ) " ) "
6671
6772pushd $script_dir /..
7277
7378QFORMAT_MODIFIED=" ${QFORMAT// ,/ _} "
7479
75- MODEL_NAME=$( basename $MODEL_PATH | sed ' s/[^0-9a-zA-Z\-]/_/g' ) _${QFORMAT_MODIFIED}${KV_CACHE_QUANT: +_kv_${KV_CACHE_QUANT} }
80+ if [ -n " $RECIPE " ]; then
81+ RECIPE_LABEL=$( basename " $RECIPE " .yaml | sed ' s/[^0-9a-zA-Z\-]/_/g' )
82+ MODEL_NAME=$( basename $MODEL_PATH | sed ' s/[^0-9a-zA-Z\-]/_/g' ) _${RECIPE_LABEL}
83+ else
84+ MODEL_NAME=$( basename $MODEL_PATH | sed ' s/[^0-9a-zA-Z\-]/_/g' ) _${QFORMAT_MODIFIED}${KV_CACHE_QUANT: +_kv_${KV_CACHE_QUANT} }
85+ fi
7686
7787SAVE_PATH=${ROOT_SAVE_PATH} /saved_models_${MODEL_NAME}
7888
@@ -181,11 +191,16 @@ if [[ $TASKS =~ "quant" ]] || [[ ! -d "$SAVE_PATH" ]] || [[ ! $(ls -A $SAVE_PATH
181191
182192 if [[ " $MODEL_CONFIG_EXIST " == false ]]; then
183193 echo " Quantizing original model..."
194+ if [ -n " $RECIPE " ]; then
195+ QUANT_ARG=" --recipe=$RECIPE "
196+ else
197+ QUANT_ARG=" --qformat=${QFORMAT// / ,} "
198+ fi
184199 python hf_ptq.py \
185200 --pyt_ckpt_path=$MODEL_PATH \
186201 --export_path=$SAVE_PATH \
187202 --sparsity_fmt=$SPARSITY_FMT \
188- --qformat= " ${QFORMAT // / ,} " \
203+ $QUANT_ARG \
189204 --calib_size=$CALIB_SIZE \
190205 --batch_size=$CALIB_BATCH_SIZE \
191206 --inference_tensor_parallel=$TP \
0 commit comments