build process

README.md

@@ -5,18 +5,30 @@ Python binding for llama.cpp using cffi
 ## Build
 
 ```bash
+#
+# setup venv
+#
 python -m venv venv
 source venv/bin/activate
 pip install poetry
-python -B examples/demo0.py
 
-huggingface-cli download bartowski/Phi-3.1-mini-4k-instruct-GGUF Phi-3.1-mini-4k-instruct-Q4_K_M.gguf
-huggingface-cli download bartowski/Phi-3.1-mini-128k-instruct-GGUF Phi-3.1-mini-128k-instruct-Q4_K_M.gguf
-huggingface-cli download IndexTeam/Index-1.9B-Chat-GGUF  ggml-model-Q4_K_M.gguf
-huggingface-cli download mradermacher/dolphin-2.9.3-qwen2-1.5b-GGUF dolphin-2.9.3-qwen2-1.5b.Q4_K_M.gguf
-huggingface-cli download mradermacher/dolphin-2.9.3-qwen2-0.5b-GGUF dolphin-2.9.3-qwen2-0.5b.Q4_K_M.gguf
-huggingface-cli download mradermacher/dolphin-2.9.3-llama-3-8b-GGUF dolphin-2.9.3-llama-3-8b.Q4_K_M.gguf
-huggingface-cli download NousResearch/Hermes-2-Pro-Llama-3-8B-GGUF Hermes-2-Pro-Llama-3-8B-Q4_K_M.gguf
+#
+# build step-by-step
+#
+# poetry run build-clean
+# poetry run build-libllama-cli-shared
+# poetry run build-libllama-cli-static
+# poetry run build-llama-cli-cffi-static
 
-CXXFLAGS="-DSHARED_LIB" LDFLAGS="-shared -o libllama-cli.so" make -j llama-cli
+#
+# build
+#
+poetry run build-all
+
+
+#
+# run demos
+#
+python -B examples/demo_cffi.py
+python -B examples/demo_ctypes.py
 ```

pyproject.toml

@@ -17,6 +17,8 @@ huggingface-hub = "^0.23.4"
 cffi = "^1.16.0"
 
 [tool.poetry.scripts]
+build-all = "scripts.build_all:build"
+build-clean = "scripts.build_clean:clean"
 build-libllama-cli-shared = "scripts.build_libllama_cli:build_shared"
 build-libllama-cli-static = "scripts.build_libllama_cli:build_static"
 build-llama-cli-cffi-static = "scripts.build_llama_cli_cffi:build_static"

scripts/build_all.py

@@ -0,0 +1,8 @@
+import subprocess
+
+
+def build():
+    subprocess.run(['poetry', 'run', 'build-clean'])
+    subprocess.run(['poetry', 'run', 'build-libllama-cli-shared'])
+    subprocess.run(['poetry', 'run', 'build-libllama-cli-static'])
+    subprocess.run(['poetry', 'run', 'build-llama-cli-cffi-static'])

scripts/build_clean.py

@@ -0,0 +1,7 @@
+import subprocess
+
+
+def clean():
+    # subprocess.run(['rm', '-v', 'llama/*.so'], check=True, shell=True)
+    subprocess.run('ls -l llama/*.so', check=True, shell=True)
+    subprocess.run(["ls", "-l", "/dev/null"], check=True, shell=True)

scripts/build_libllama_cli.py

@@ -8,22 +8,20 @@ def build_shared():
     env['CXXFLAGS'] = '-DSHARED_LIB'
     env['LDFLAGS'] = '-shared -o libllama-cli.so'
 
-    subprocess.run(['rm', '-rf', 'llama.cpp'])
-    subprocess.run(['rm', '-rf', 'llama/libllama-cli.so'])
-    subprocess.run(['git', 'clone', 'https://github.com/ggerganov/llama.cpp.git'])
-    subprocess.run(['patch', 'llama.cpp/examples/main/main.cpp', 'main_shared_library_0.patch'])
+    subprocess.run(['rm', '-rf', 'llama.cpp'], check=True)
+    subprocess.run(['git', 'clone', 'https://github.com/ggerganov/llama.cpp.git'], check=True)
+    subprocess.run(['patch', 'llama.cpp/examples/main/main.cpp', 'main_shared_library_0.patch'], check=True)
     subprocess.run(['make', '-C', 'llama.cpp', '-j', 'llama-cli'], check=True, env=env)
 
-    shutil.copy("llama.cpp/libllama-cli.so", "llama/libllama-cli.so")
+    shutil.copy("llama.cpp/libllama-cli.so", "llama/libllama-cli.so", check=True)
 
 
 def build_static():
     env = os.environ.copy()
     env['CXXFLAGS'] = '-DSHARED_LIB'
 
-    subprocess.run(['rm', '-rf', 'llama.cpp'])
-    subprocess.run(['rm', '-rf', 'llama/libllama-cli.a'])
-    subprocess.run(['git', 'clone', 'https://github.com/ggerganov/llama.cpp.git'])
-    subprocess.run(['patch', 'llama.cpp/examples/main/main.cpp', 'main_shared_library_0.patch'])
-    subprocess.run(['patch', 'llama.cpp/Makefile', 'makefile_static_library_0.patch'])
+    subprocess.run(['rm', '-rf', 'llama.cpp'], check=True)
+    subprocess.run(['git', 'clone', 'https://github.com/ggerganov/llama.cpp.git'], check=True)
+    subprocess.run(['patch', 'llama.cpp/examples/main/main.cpp', 'main_shared_library_0.patch'], check=True)
+    subprocess.run(['patch', 'llama.cpp/Makefile', 'makefile_static_library_0.patch'], check=True)
     subprocess.run(['make', '-C', 'llama.cpp', '-j', 'llama-cli-a', 'GGML_NO_OPENMP=1', 'GGML_NO_LLAMAFILE=1'], check=True, env=env)

scripts/build_llama_cli_cffi.py

@@ -31,6 +31,9 @@
 
 
 def build_static():
+    subprocess.run(['rm', '-rf', 'llama.cpp'], check=True)
+    subprocess.run(['git', 'clone', 'https://github.com/ggerganov/llama.cpp.git'], check=True)
+
     ffibuilder.compile(tmpdir="build", verbose=True)
 
     for file in glob.glob('build/*.so'):