Merge branch 'main' of github.com:abetlen/llama_cpp_python into main

llama_cpp/llama.py

@@ -238,6 +238,7 @@ class Llama:
             n_ctx: Maximum context size.
             n_parts: Number of parts to split the model into. If -1, the number of parts is automatically determined.
             seed: Random seed. -1 for random.
+            n_gpu_layers: Number of layers to offload to GPU (-ngl). If -1, all layers are offloaded.
             f16_kv: Use half-precision for key/value cache.
             logits_all: Return logits for all tokens, not just the last token.
             vocab_only: Only load the vocabulary no weights.
@@ -266,7 +267,7 @@ class Llama:
 
         self.params = llama_cpp.llama_context_default_params()
         self.params.n_ctx = n_ctx
-        self.params.n_gpu_layers = n_gpu_layers
+        self.params.n_gpu_layers = 0x7FFFFFFF if n_gpu_layers == -1 else n_gpu_layers  # 0x7FFFFFFF is INT32 max, will be auto set to all layers
         self.params.seed = seed
         self.params.f16_kv = f16_kv
         self.params.logits_all = logits_all

setup.py

@@ -15,6 +15,7 @@ setup(
     author_email="abetlen@gmail.com",
     license="MIT",
     package_dir={"llama_cpp": "llama_cpp", "llama_cpp.server": "llama_cpp/server"},
+    package_data={"llama_cpp": ["py.typed"]},
     packages=["llama_cpp", "llama_cpp.server"],
     install_requires=["typing-extensions>=4.5.0", "numpy>=1.20.0", "diskcache>=5.6.1"],
     extras_require={