created a pip package

README.md

@@ -15,7 +15,7 @@ This project aims to address the third using LLaMa.cpp and GGML.
 
 - Inference Speed! Focus on inference, not training.
 - Precompressed models.
-- Minimal setup required - soon `pip install cformers` should be good to get started.
+- Minimal setup required - `pip install cformers` should be good to get started.
 - Easily switch between models and quantization types.
 - Support variety of prompts.
 
@@ -26,14 +26,12 @@ And most importantly:
 
 Setup
 ```bash
-pip install transformers wget
-git clone https://github.com/nolanoOrg/cformers.git
-cd cformers/cformers/cpp && make && cd ..
+pip install cformers
 ```
 
 Usage:
 ```python
-from interface import AutoInference as AI
+from cformers import AutoInference as AI
 ai = AI('EleutherAI/gpt-j-6B')
 x = ai.generate('def parse_html(html_doc):', num_tokens_to_generate=500)
 print(x['token_str'])
@@ -58,8 +56,6 @@ chat.py accepts the following parameteres:
 - ```-p Tell me a joke``` for a single prompt interaction
 - ```-m pythia``` to load one of the available (bloom, pythia or gptj )
 
-We are working on adding support for `pip install cformers.`
-
 Following Architectures are supported:
 - GPT-J
 - BLOOM

cformers/__init__.py

@@ -1,2 +1,2 @@
 """Cformers: SoTA Transformer inference on your CPU."""
-from .interface import AutoModel, AutoTokenizer
+from .interface import AutoInference

cformers/interface.py

@@ -7,6 +7,7 @@
 import select
 import wget
 import requests
+import pathlib
 
 import transformers as tf # RIP TensorFlow
 
@@ -193,11 +194,13 @@ def generate(self,
             f"Prompt should be a list of integers {prompt}"
         # Convert to a string of space separated integers
         prompt = " ".join([str(x) for x in prompt])
+
+        main_file = str(pathlib.Path(__file__).parent.resolve())
 
         if os.name == 'nt':
-            main_file = "./cpp/main.exe"
+            main_file += "/cpp/main.exe"
         else:
-            main_file = "./cpp/main"
+            main_file += "/cpp/main"
 
         command = [main_file, self.cpp_model_name,
                    "-m", self.model_save_path,

setup.py

@@ -0,0 +1,40 @@
+from setuptools import setup, find_packages
+import codecs
+import os
+import subprocess
+
+packages= ['cformers', 'cformers/cpp']
+package_data = {'cformers': ['*'], 'cformers/cpp': ['*']}
+build_main = subprocess.run(["make"], stdout=subprocess.PIPE, cwd="cformers/cpp")
+
+here = os.path.abspath(os.path.dirname(__file__))
+
+with codecs.open(os.path.join(here, "README.md"), encoding="utf-8") as fh:
+    long_description = "\n" + fh.read()
+
+VERSION = '0.0.4'
+DESCRIPTION = 'SoTA Transformers with C-backend for fast inference on your CPU.'
+LONG_DESCRIPTION = 'We identify three pillers to enable fast inference of SoTA AI models on your CPU:\n1. Fast C/C++ LLM inference kernels for CPU.\n2. Machine Learning Research & Exploration front - Compression through quantization, sparsification, training on more data, collecting data and training instruction & chat models.\n3. Easy to use API for fast AI inference in dynamically typed language like Python.\n\nThis project aims to address the third using LLaMa.cpp and GGML.'
+
+# Setting up
+setup(
+    name="cformers",
+    version=VERSION,
+    author="Ayush Kaushal (Ayushk4)",
+    author_email="[email protected]",
+    description=DESCRIPTION,
+    long_description_content_type="text/markdown",
+    long_description=LONG_DESCRIPTION,
+    packages=packages,
+    package_data=package_data,
+    install_requires=['transformers', 'torch', 'wget'],
+    keywords=['python', 'local inference', 'c++ inference', 'language models', 'cpu inference', 'quantization'],
+    classifiers=[
+        "Development Status :: 2 - Pre-Alpha",
+        "Intended Audience :: Developers",
+        "Programming Language :: Python :: 3",
+        "Operating System :: Unix",
+        "Operating System :: MacOS :: MacOS X",
+        "Operating System :: Microsoft :: Windows",
+    ]
+)