model-conversion : use BUILD_DIR variable in all scripts (#19015)

This commit modifies all the utility scripts to use an optional
BUILD_DIR variable/argument to specify the build directory.

The motivation for this is that Commit
3d55846a5c626e2e608db8c24fa9ee6defaacca9 ("model-conversion : add
BUILD_DIR variable to run-converted-model scripts") introduced this
variable to the causal and embeddings scripts, but I missed the scripts
in the utils directory.

examples/model-conversion/scripts/utils/perplexity-gen.sh

@@ -3,6 +3,7 @@
 set -e
 
 CONVERTED_MODEL="${1:-"$CONVERTED_MODEL"}"
+BUILD_DIR="${2:-"$BUILD_DIR"}"
 
 # Final check if we have a model path
 if [ -z "$CONVERTED_MODEL" ]; then
@@ -25,9 +26,13 @@ mkdir -p ppl
 OUTPUTFILE="ppl/$(basename $CONVERTED_MODEL).kld"
 echo "Model: $CONVERTED_MODEL"
 
-cmake --build ../../build --target llama-perplexity -j8
+if [ -z "$BUILD_DIR" ]; then
+    BUILD_DIR="../../build"
+fi
+
+cmake --build $BUILD_DIR --target llama-perplexity -j8
 
-../.././build/bin/llama-perplexity -m $CONVERTED_MODEL \
+${BUILD_DIR}/bin/llama-perplexity -m $CONVERTED_MODEL \
     -f ppl/wikitext-2-raw/wiki.test.raw \
     --kl-divergence-base $OUTPUTFILE
 

examples/model-conversion/scripts/utils/perplexity-run-simple.sh

@@ -3,6 +3,7 @@
 set -e
 
 QUANTIZED_MODEL="${1:-"$QUANTIZED_MODEL"}"
+BUILD_DIR="${2:-"$BUILD_DIR"}"
 
 if [ -z "$QUANTIZED_MODEL" ]; then
     echo "Error: Model path must be provided either as:" >&2
@@ -20,8 +21,12 @@ if [ ! -d "ppl/wikitext-2-raw" ]; then
     popd
 fi
 
-cmake --build ../../build --target llama-perplexity -j8
+if [ -z "$BUILD_DIR" ]; then
+    BUILD_DIR="../../build"
+fi
+
+cmake --build $BUILD_DIR --target llama-perplexity -j8
 
-../.././build/bin/llama-perplexity -m $QUANTIZED_MODEL -f ppl/wikitext-2-raw/wiki.test.raw
+${BUILD_DIR}/bin/llama-perplexity -m $QUANTIZED_MODEL -f ppl/wikitext-2-raw/wiki.test.raw
 
 

examples/model-conversion/scripts/utils/perplexity-run.sh

@@ -3,7 +3,8 @@
 set -e
 
 QUANTIZED_MODEL="${1:-"$QUANTIZED_MODEL"}"
-LOGITS_FILE="${1:-"$LOGITS_FILE"}"
+LOGITS_FILE="${2:-"$LOGITS_FILE"}"
+BUILD_DIR="${3:-"$BUILD_DIR"}"
 
 if [ -z "$QUANTIZED_MODEL" ]; then
     echo "Error: Model path must be provided either as:" >&2
@@ -18,11 +19,15 @@ if [ ! -f ${LOGITS_FILE} ]; then
     exit 1
 fi
 
+if [ -z "$BUILD_DIR" ]; then
+    BUILD_DIR="../../build"
+fi
+
 echo "Model: $QUANTIZED_MODEL"
 echo "Data file: $LOGITS_FILE"
 
-cmake --build ../../build --target llama-perplexity -j8
+cmake --build $BUILD_DIR --target llama-perplexity -j8
 
-../.././build/bin/llama-perplexity -m $QUANTIZED_MODEL \
+${BUILD_DIR}/bin/llama-perplexity -m $QUANTIZED_MODEL \
     --kl-divergence-base $LOGITS_FILE \
     --kl-divergence

examples/model-conversion/scripts/utils/quantize.sh

@@ -6,6 +6,7 @@ CONVERTED_MODEL="${1:-"$CONVERTED_MODEL"}"
 QUANTIZED_TYPE="${2:-"$QUANTIZED_TYPE"}"
 TOKEN_EMBD_TYPE="${3:-"${TOKEN_EMBD_TYPE}"}"
 OUTPUT_TYPE="${4:-"${OUTPUT_TYPE}"}"
+BUILD_DIR="${5:-"$BUILD_DIR"}"
 QUANTIZED_MODEL=$CONVERTED_MODEL
 
 # Final check if we have a model path
@@ -33,12 +34,16 @@ else
     exit 1
 fi
 
-cmake --build ../../build --target llama-quantize -j8
+if [ -z "$BUILD_DIR" ]; then
+    BUILD_DIR="../../build"
+fi
+
+cmake --build $BUILD_DIR --target llama-quantize -j8
 
 echo $TOKEN_EMBD_TYPE
 echo $OUTPUT_TYPE
 
-CMD_ARGS=("../../build/bin/llama-quantize")
+CMD_ARGS=("${BUILD_DIR}/bin/llama-quantize")
 [[ -n "$TOKEN_EMBD_TYPE" ]] && CMD_ARGS+=("--token-embedding-type" "$TOKEN_EMBD_TYPE")
 [[ -n "$OUTPUT_TYPE" ]]     && CMD_ARGS+=("--output-tensor-type" "$OUTPUT_TYPE")
 CMD_ARGS+=("$CONVERTED_MODEL" "$QUANTIZED_MODEL" "$QUANTIZED_TYPE")

examples/model-conversion/scripts/utils/run-embedding-server.sh

@@ -4,6 +4,7 @@ set -e
 #
 # First try command line argument, then environment variable, then file
 CONVERTED_MODEL="${1:-"$CONVERTED_MODEL"}"
+BUILD_DIR="${2:-"$BUILD_DIR"}"
 
 # Final check if we have a model path
 if [ -z "$CONVERTED_MODEL" ]; then
@@ -13,10 +14,14 @@ if [ -z "$CONVERTED_MODEL" ]; then
     exit 1
 fi
 
+if [ -z "$BUILD_DIR" ]; then
+    BUILD_DIR="../../build"
+fi
+
 echo $CONVERTED_MODEL
 
-cmake --build ../../build --target llama-server
+cmake --build $BUILD_DIR --target llama-server
 
-../../build/bin/llama-server -m $CONVERTED_MODEL \
+${BUILD_DIR}/bin/llama-server -m $CONVERTED_MODEL \
     --embedding \
     --pooling none