Update app.py

app.py

@@ -47,11 +47,11 @@ def setup_llama_cpp():
             raise RuntimeError(f"Failed to install llama.cpp requirements. Error: {e.stderr}")
 
     if not LLAMA_CPP_QUANTIZE_SCRIPT.exists():
-        logging.info("llama.cpp 'quantize' binary not found. Building with CMake...")
+        logging.info("llama.cpp binaries not found. Building with CMake...")
         try:
             subprocess.run(["cmake", "."], cwd=str(LLAMA_CPP_DIR), check=True, capture_output=True, text=True)
-            subprocess.run(["cmake", "--build", ".", "--target", "quantize"], cwd=str(LLAMA_CPP_DIR), check=True, capture_output=True, text=True)
-            logging.info("'quantize' binary built successfully with CMake.")
+            subprocess.run(["cmake", "--build", "."], cwd=str(LLAMA_CPP_DIR), check=True, capture_output=True, text=True)
+            logging.info("llama.cpp binaries built successfully with CMake.")
         except subprocess.CalledProcessError as e:
             raise RuntimeError(f"Failed to build llama.cpp with CMake. Error: {e.stderr}")
 
@@ -121,12 +121,13 @@ def stage_3_4_gguf_quantize(model_path_or_id: str, original_model_id: str, quant
     model_name = original_model_id.replace('/', '_')
     gguf_path = os.path.join(OUTPUT_DIR, f"{model_name}-{run_id}-gguf")
     os.makedirs(gguf_path, exist_ok=True)
-    f16_gguf_path = os.path.join(gguf_path, "model-f16.gguf")
-    quantized_gguf_path = os.path.join(gguf_path, "model.gguf")
+    f16_gguf_path = os.path.abspath(os.path.join(gguf_path, "model-f16.gguf"))
+    quantized_gguf_path = os.path.abspath(os.path.join(gguf_path, "model.gguf"))
+    absolute_model_path = os.path.abspath(model_path_or_id) if os.path.exists(model_path_or_id) else model_path_or_id
     try:
-        convert_command = ["python3", "convert.py", model_path_or_id, "--outfile", f16_gguf_path, "--outtype", "f16"]
+        convert_command = ["python3", "convert.py", absolute_model_path, "--outfile", f16_gguf_path, "--outtype", "f16"]
         process = subprocess.run(convert_command, check=True, capture_output=True, text=True, cwd=str(LLAMA_CPP_DIR))
-        log_stream += f"Executing llama.cpp conversion script on '{model_path_or_id}'...\n{process.stdout}\n"
+        log_stream += f"Executing llama.cpp conversion script...\n{process.stdout}\n"
         if process.stderr: log_stream += f"[STDERR]\n{process.stderr}\n"
         quantize_map = {"q4_k_m": "Q4_K_M", "q5_k_m": "Q5_K_M", "q8_0": "Q8_0", "f16": "F16"}
         target_quant_name = quantize_map.get(quantization_strategy.lower(), "Q4_K_M")