diff --git a/python/llm/src/bigdl/llm/ggml/model/bloom/bloom_cpp.py b/python/llm/src/bigdl/llm/ggml/model/bloom/bloom_cpp.py index acde790b..16824654 100644 --- a/python/llm/src/bigdl/llm/ggml/model/bloom/bloom_cpp.py +++ b/python/llm/src/bigdl/llm/ggml/model/bloom/bloom_cpp.py @@ -81,6 +81,7 @@ def _load_shared_library(lib_base_name: str): # Add the library directory to the DLL search path on Windows (if needed) if sys.platform == "win32" and sys.version_info >= (3, 8): os.add_dll_directory(str(_base_path)) + os.environ["PATH"] = str(_base_path) + ";" + os.environ["PATH"] # Try to load the shared library, handling potential errors for _lib_path in _lib_paths: diff --git a/python/llm/src/bigdl/llm/ggml/model/gptneox/gptneox_cpp.py b/python/llm/src/bigdl/llm/ggml/model/gptneox/gptneox_cpp.py index 30577624..d7d9b26c 100644 --- a/python/llm/src/bigdl/llm/ggml/model/gptneox/gptneox_cpp.py +++ b/python/llm/src/bigdl/llm/ggml/model/gptneox/gptneox_cpp.py @@ -81,6 +81,7 @@ def _load_shared_library(lib_base_name: str): # Add the library directory to the DLL search path on Windows (if needed) if sys.platform == "win32" and sys.version_info >= (3, 8): os.add_dll_directory(str(_base_path)) + os.environ["PATH"] = str(_base_path) + ";" + os.environ["PATH"] cdll_args["winmode"] = 0 # Try to load the shared library, handling potential errors diff --git a/python/llm/src/bigdl/llm/ggml/model/llama/llama_cpp.py b/python/llm/src/bigdl/llm/ggml/model/llama/llama_cpp.py index f7e84321..bb456c9b 100644 --- a/python/llm/src/bigdl/llm/ggml/model/llama/llama_cpp.py +++ b/python/llm/src/bigdl/llm/ggml/model/llama/llama_cpp.py @@ -81,6 +81,7 @@ def _load_shared_library(lib_base_name: str): # Add the library directory to the DLL search path on Windows (if needed) if sys.platform == "win32" and sys.version_info >= (3, 8): os.add_dll_directory(str(_base_path)) + os.environ["PATH"] = str(_base_path) + ";" + os.environ["PATH"] if "CUDA_PATH" in os.environ: os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "bin")) os.add_dll_directory(os.path.join(os.environ["CUDA_PATH"], "lib")) diff --git a/python/llm/src/bigdl/llm/ggml/model/starcoder/starcoder_cpp.py b/python/llm/src/bigdl/llm/ggml/model/starcoder/starcoder_cpp.py index f1b010b1..7f50d0fa 100644 --- a/python/llm/src/bigdl/llm/ggml/model/starcoder/starcoder_cpp.py +++ b/python/llm/src/bigdl/llm/ggml/model/starcoder/starcoder_cpp.py @@ -81,6 +81,7 @@ def _load_shared_library(lib_base_name: str): # Add the library directory to the DLL search path on Windows (if needed) if sys.platform == "win32" and sys.version_info >= (3, 8): os.add_dll_directory(str(_base_path)) + os.environ["PATH"] = str(_base_path) + ";" + os.environ["PATH"] # Try to load the shared library, handling potential errors for _lib_path in _lib_paths: diff --git a/python/llm/src/bigdl/llm/utils/utils.py b/python/llm/src/bigdl/llm/utils/utils.py index 996ff6d2..8c0b5721 100644 --- a/python/llm/src/bigdl/llm/utils/utils.py +++ b/python/llm/src/bigdl/llm/utils/utils.py @@ -30,7 +30,8 @@ def get_cpu_flags(): else: invalidOperationError(False, "Unsupported CPUFLAGS.") else: - flags = "_vnni" if check_avx_vnni() else "" + # flags = "_vnni" if check_avx_vnni() else "" + flags = "-api" return flags