Upload scripts/quantizer.sh with huggingface_hub

scripts/quantizer.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+set -euo pipefail
+
+# Simple quantization script for llama.cpp models
+# Usage: ./quantizer.sh <imatrix_file> <input_model> [quant_type]
+
+IMATRIX="${1:?Error: imatrix file required}"
+INPUT_MODEL="${2:?Error: input model required}"
+QUANT_TYPE="${3:-Q4_K_M}"
+
+# Extract base name for output
+BASE_NAME=$(basename "$INPUT_MODEL" .gguf)
+OUTPUT_MODEL="${BASE_NAME}-${QUANT_TYPE}.gguf"
+
+# Find llama-quantize binary
+if command -v llama-quantize &>/dev/null; then
+    QUANTIZE_BIN="llama-quantize"
+elif [ -f "$HOME/llama.cpp/build/bin/llama-quantize" ]; then
+    QUANTIZE_BIN="$HOME/llama.cpp/build/bin/llama-quantize"
+else
+    echo "Error: llama-quantize not found"
+    echo "Install llama.cpp or add it to PATH"
+    exit 1
+fi
+
+echo "Quantizing model..."
+echo "  Input:  $INPUT_MODEL"
+echo "  Output: $OUTPUT_MODEL"
+echo "  Type:   $QUANT_TYPE"
+echo "  Imatrix: $IMATRIX"
+echo ""
+
+"$QUANTIZE_BIN" \
+    --imatrix "$IMATRIX" \
+    "$INPUT_MODEL" \
+    "$OUTPUT_MODEL" \
+    "$QUANT_TYPE"
+
+echo ""
+echo "Done! Output: $OUTPUT_MODEL"