diff --git a/neuralcoref/cli/include/msvc9/stdint.h b/include/msvc9/stdint.h
similarity index 100%
rename from neuralcoref/cli/include/msvc9/stdint.h
rename to include/msvc9/stdint.h
diff --git a/include/murmurhash/MurmurHash2.h b/include/murmurhash/MurmurHash2.h
new file mode 100644
index 0000000..6d7ccf4
--- /dev/null
+++ b/include/murmurhash/MurmurHash2.h
@@ -0,0 +1,22 @@
+//-----------------------------------------------------------------------------
+// MurmurHash2 was written by Austin Appleby, and is placed in the public
+// domain. The author hereby disclaims copyright to this source code.
+
+#ifndef _MURMURHASH2_H_
+#define _MURMURHASH2_H_
+
+#include <stdint.h>
+
+//-----------------------------------------------------------------------------
+
+uint32_t MurmurHash2        ( const void * key, int len, uint32_t seed );
+uint64_t MurmurHash64A      ( const void * key, int len, uint64_t seed );
+uint64_t MurmurHash64B      ( const void * key, int len, uint64_t seed );
+uint32_t MurmurHash2A       ( const void * key, int len, uint32_t seed );
+uint32_t MurmurHashNeutral2 ( const void * key, int len, uint32_t seed );
+uint32_t MurmurHashAligned2 ( const void * key, int len, uint32_t seed );
+
+//-----------------------------------------------------------------------------
+
+#endif // _MURMURHASH2_H_
+
diff --git a/include/murmurhash/MurmurHash3.h b/include/murmurhash/MurmurHash3.h
new file mode 100644
index 0000000..9b4c3c9
--- /dev/null
+++ b/include/murmurhash/MurmurHash3.h
@@ -0,0 +1,28 @@
+//-----------------------------------------------------------------------------
+// MurmurHash3 was written by Austin Appleby, and is placed in the public
+// domain. The author hereby disclaims copyright to this source code.
+
+#ifndef _MURMURHASH3_H_
+#define _MURMURHASH3_H_
+
+#include <stdint.h>
+
+//-----------------------------------------------------------------------------
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+void MurmurHash3_x86_32  ( const void * key, int len, uint32_t seed, void * out );
+
+void MurmurHash3_x86_128 ( const void * key, int len, uint32_t seed, void * out );
+
+void MurmurHash3_x64_128 ( const void * key, int len, uint32_t seed, void * out );
+
+#ifdef __cplusplus
+}
+#endif
+
+//-----------------------------------------------------------------------------
+
+#endif // _MURMURHASH3_H_
diff --git a/neuralcoref/cli/include/numpy/__multiarray_api.h b/include/numpy/__multiarray_api.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/__multiarray_api.h
rename to include/numpy/__multiarray_api.h
diff --git a/neuralcoref/cli/include/numpy/__ufunc_api.h b/include/numpy/__ufunc_api.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/__ufunc_api.h
rename to include/numpy/__ufunc_api.h
diff --git a/neuralcoref/cli/include/numpy/_neighborhood_iterator_imp.h b/include/numpy/_neighborhood_iterator_imp.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/_neighborhood_iterator_imp.h
rename to include/numpy/_neighborhood_iterator_imp.h
diff --git a/neuralcoref/cli/include/numpy/_numpyconfig.h b/include/numpy/_numpyconfig.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/_numpyconfig.h
rename to include/numpy/_numpyconfig.h
diff --git a/neuralcoref/cli/include/numpy/arrayobject.h b/include/numpy/arrayobject.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/arrayobject.h
rename to include/numpy/arrayobject.h
diff --git a/neuralcoref/cli/include/numpy/arrayscalars.h b/include/numpy/arrayscalars.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/arrayscalars.h
rename to include/numpy/arrayscalars.h
diff --git a/neuralcoref/cli/include/numpy/halffloat.h b/include/numpy/halffloat.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/halffloat.h
rename to include/numpy/halffloat.h
diff --git a/neuralcoref/cli/include/numpy/multiarray_api.txt b/include/numpy/multiarray_api.txt
similarity index 100%
rename from neuralcoref/cli/include/numpy/multiarray_api.txt
rename to include/numpy/multiarray_api.txt
diff --git a/neuralcoref/cli/include/numpy/ndarrayobject.h b/include/numpy/ndarrayobject.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/ndarrayobject.h
rename to include/numpy/ndarrayobject.h
diff --git a/neuralcoref/cli/include/numpy/ndarraytypes.h b/include/numpy/ndarraytypes.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/ndarraytypes.h
rename to include/numpy/ndarraytypes.h
diff --git a/neuralcoref/cli/include/numpy/noprefix.h b/include/numpy/noprefix.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/noprefix.h
rename to include/numpy/noprefix.h
diff --git a/neuralcoref/cli/include/numpy/npy_3kcompat.h b/include/numpy/npy_3kcompat.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_3kcompat.h
rename to include/numpy/npy_3kcompat.h
diff --git a/neuralcoref/cli/include/numpy/npy_common.h b/include/numpy/npy_common.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_common.h
rename to include/numpy/npy_common.h
diff --git a/neuralcoref/cli/include/numpy/npy_cpu.h b/include/numpy/npy_cpu.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_cpu.h
rename to include/numpy/npy_cpu.h
diff --git a/neuralcoref/cli/include/numpy/npy_deprecated_api.h b/include/numpy/npy_deprecated_api.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_deprecated_api.h
rename to include/numpy/npy_deprecated_api.h
diff --git a/neuralcoref/cli/include/numpy/npy_endian.h b/include/numpy/npy_endian.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_endian.h
rename to include/numpy/npy_endian.h
diff --git a/neuralcoref/cli/include/numpy/npy_interrupt.h b/include/numpy/npy_interrupt.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_interrupt.h
rename to include/numpy/npy_interrupt.h
diff --git a/neuralcoref/cli/include/numpy/npy_math.h b/include/numpy/npy_math.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_math.h
rename to include/numpy/npy_math.h
diff --git a/neuralcoref/cli/include/numpy/npy_no_deprecated_api.h b/include/numpy/npy_no_deprecated_api.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_no_deprecated_api.h
rename to include/numpy/npy_no_deprecated_api.h
diff --git a/neuralcoref/cli/include/numpy/npy_os.h b/include/numpy/npy_os.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/npy_os.h
rename to include/numpy/npy_os.h
diff --git a/neuralcoref/cli/include/numpy/numpyconfig.h b/include/numpy/numpyconfig.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/numpyconfig.h
rename to include/numpy/numpyconfig.h
diff --git a/neuralcoref/cli/include/numpy/old_defines.h b/include/numpy/old_defines.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/old_defines.h
rename to include/numpy/old_defines.h
diff --git a/neuralcoref/cli/include/numpy/oldnumeric.h b/include/numpy/oldnumeric.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/oldnumeric.h
rename to include/numpy/oldnumeric.h
diff --git a/neuralcoref/cli/include/numpy/ufunc_api.txt b/include/numpy/ufunc_api.txt
similarity index 100%
rename from neuralcoref/cli/include/numpy/ufunc_api.txt
rename to include/numpy/ufunc_api.txt
diff --git a/neuralcoref/cli/include/numpy/ufuncobject.h b/include/numpy/ufuncobject.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/ufuncobject.h
rename to include/numpy/ufuncobject.h
diff --git a/neuralcoref/cli/include/numpy/utils.h b/include/numpy/utils.h
similarity index 100%
rename from neuralcoref/cli/include/numpy/utils.h
rename to include/numpy/utils.h
diff --git a/neuralcoref/__init__.py b/neuralcoref/__init__.py
index 79caa3e..fac4887 100644
--- a/neuralcoref/__init__.py
+++ b/neuralcoref/__init__.py
@@ -1,27 +1,69 @@
 # coding: utf8
-from __future__ import unicode_literals
+"""
+Simple check list from AllenNLP repo: https://github.com/allenai/allennlp/blob/master/setup.py
 
+To create the package for pypi.
+
+1. Change the version in __init__.py and setup.py.
+
+2. Commit these changes with the message: "Release: VERSION"
+
+3. Add a tag in git to mark the release: "git tag VERSION -m'Adds tag VERSION for pypi' "
+   Push the tag to git: git push --tags origin master
+
+4. Build both the sources and the wheel. Do not change anything in setup.py between
+   creating the wheel and the source distribution (obviously).
+
+   For the wheel, run: "python setup.py bdist_wheel" in the top level allennlp directory.
+   (this will build a wheel for the python version you use to build it - make sure you use python 3.x).
+
+   For the sources, run: "python setup.py sdist"
+   You should now have a /dist directory with both .whl and .tar.gz source versions of allennlp.
+
+5. Check that everything looks correct by uploading the package to the pypi test server:
+
+   twine upload dist/* -r pypitest
+   (pypi suggest using twine as other methods upload files via plaintext.)
+
+   Check that you can install it in a virtualenv by running:
+   pip install -i https://testpypi.python.org/pypi neuralcoref
+
+6. Upload the final version to actual pypi:
+   twine upload dist/* -r pypi
+
+7. Copy the release notes from RELEASE.md to the tag in github once everything is looking hunky-dory.
+
+"""
+from __future__ import unicode_literals, absolute_import
+
+import os
 import shutil
 import tarfile
 import tempfile
 import logging
 
 from .neuralcoref import NeuralCoref
-from .file_utils import NEURALCOREF_CACHE, cached_path
+from .file_utils import NEURALCOREF_MODEL_URL, NEURALCOREF_MODEL_PATH, NEURALCOREF_CACHE, cached_path
 
-__all__ = ['NeuralCoref']
+__all__ = ['NeuralCoref', 'add_to_pipe']
+__version__ = "4.0.0"
 
 logger = logging.getLogger(__name__)
 
-MODEL_URL = "https://s3.amazonaws.com/models.huggingface.co/neuralcoref/neuralcoref_model.tar.gz"
-LOCAL_PATH = os.path.join(str(NEURALCOREF_CACHE), "/neuralcoref/")
-
-try:
-    local_model = cached_path(LOCAL_PATH)
-except:
-    os.makedirs(LOCAL_PATH)
-    downloaded_model = cached_path(MODEL_URL)
+if os.path.exists(NEURALCOREF_MODEL_PATH) and os.path.exists(os.path.join(NEURALCOREF_MODEL_PATH, "cfg")):
+    logger.info("Loading model from {}".format(NEURALCOREF_MODEL_PATH))
+    local_model = cached_path(NEURALCOREF_MODEL_PATH)
+else:
+    if not os.path.exists(NEURALCOREF_MODEL_PATH):
+        os.makedirs(NEURALCOREF_MODEL_PATH)
+    logger.info("Getting model from {} or cache".format(NEURALCOREF_MODEL_URL))
+    downloaded_model = cached_path(NEURALCOREF_MODEL_URL)
 
-    logger.info("extracting archive file {} to dir {}".format(downloaded_model, LOCAL_PATH))
+    logger.info("extracting archive file {} to dir {}".format(downloaded_model, NEURALCOREF_MODEL_PATH))
     with tarfile.open(downloaded_model, 'r:gz') as archive:
-        archive.extractall(LOCAL_PATH)
+        archive.extractall(NEURALCOREF_CACHE)
+
+def add_to_pipe(nlp):
+    coref = NeuralCoref(nlp.vocab)
+    nlp.add_pipe(coref, name='neuralcoref')
+    return nlp
diff --git a/neuralcoref/cli/__init__.py b/neuralcoref/cli/__init__.py
deleted file mode 100644
index b6ac118..0000000
--- a/neuralcoref/cli/__init__.py
+++ /dev/null
@@ -1,2 +0,0 @@
-from .package import package
-from .download import download
\ No newline at end of file
diff --git a/neuralcoref/cli/download.py b/neuralcoref/cli/download.py
deleted file mode 100644
index bd43da8..0000000
--- a/neuralcoref/cli/download.py
+++ /dev/null
@@ -1,86 +0,0 @@
-# coding: utf8
-# Copied from spaCy repo - all glory to Explosion AI
-from __future__ import unicode_literals
-
-import plac
-import requests
-import os
-import subprocess
-import sys
-
-from spacy.cli._messages import Messages
-from spacy.util import prints, get_package_path
-from spacy import util
-from .. import about
-
-@plac.annotations(
-    model=("model to download, shortcut or name)", "positional", None, str),
-    direct=("force direct download. Needs model name with version and won't "
-            "perform compatibility check", "flag", "d", bool),
-    pip_args=("additional arguments to be passed to `pip install` when "
-              "installing the model"))
-def download(model, direct=False, *pip_args):
-    """
-    Download compatible model from default download path using pip. Model
-    can be shortcut, model name or, if --direct flag is set, full model name
-    with version.
-    """
-    if direct:
-        dl = download_model('{m}/{m}.tar.gz#egg={m}'.format(m=model), pip_args)
-    else:
-        shortcuts = get_json(about.__shortcuts__, "available shortcuts")
-        model_name = shortcuts.get(model, model)
-        compatibility = get_compatibility()
-        version = get_version(model_name, compatibility)
-        dl = download_model('{m}-{v}/{m}-{v}.tar.gz#egg={m}=={v}'
-                            .format(m=model_name, v=version), pip_args)
-        if dl != 0:  # if download subprocess doesn't return 0, exit
-            sys.exit(dl)
-        try:
-            # Get package path here because link uses
-            # pip.get_installed_distributions() to check if model is a
-            # package, which fails if model was just installed via
-            # subprocess
-            package_path = get_package_path(model_name)
-            link(model_name, model, force=True, model_path=package_path)
-        except:
-            # Dirty, but since neuralcoref.download and the auto-linking is
-            # mostly a convenience wrapper, it's best to show a success
-            # message and loading instructions, even if linking fails.
-            prints(Messages.M001.format(name=model_name), title=Messages.M002)
-
-
-def get_json(url, desc):
-    r = requests.get(url)
-    if r.status_code != 200:
-        prints(Messages.M004.format(desc=desc, version=about.__version__),
-               title=Messages.M003.format(code=r.status_code), exits=1)
-    return r.json()
-
-
-def get_compatibility():
-    version = about.__version__
-    version = version.rsplit('.dev', 1)[0]
-    comp_table = get_json(about.__compatibility__, "compatibility table")
-    comp = comp_table['neuralcoref']
-    if version not in comp:
-        prints(Messages.M006.format(version=version), title=Messages.M005,
-               exits=1)
-    return comp[version]
-
-
-def get_version(model, comp):
-    model = model.rsplit('.dev', 1)[0]
-    if model not in comp:
-        prints(Messages.M007.format(name=model, version=about.__version__),
-               title=Messages.M005, exits=1)
-    return comp[model][0]
-
-
-def download_model(filename, user_pip_args=None):
-    download_url = about.__download_url__ + '/' + filename
-    pip_args = ['--no-cache-dir', '--no-deps']
-    if user_pip_args:
-        pip_args.extend(user_pip_args)
-    cmd = [sys.executable, '-m', 'pip', 'install'] + pip_args + [download_url]
-    return subprocess.call(cmd, env=os.environ.copy())
diff --git a/neuralcoref/cli/package.py b/neuralcoref/cli/package.py
deleted file mode 100644
index 41e9333..0000000
--- a/neuralcoref/cli/package.py
+++ /dev/null
@@ -1,308 +0,0 @@
-# coding: utf8
-# Copied from spaCy repo - all glory to Explosion AI
-from __future__ import unicode_literals
-
-import shutil
-from pathlib import Path
-import plac
-
-from srsly import json_dumps, read_json
-
-from spacy.cli._messages import Messages
-from spacy.compat import path2str
-from spacy.util import prints
-from spacy import util
-from spacy import about
-
-
-@plac.annotations(
-    input_dir=("directory with model data", "positional", None, str),
-    output_dir=("output parent directory", "positional", None, str),
-    meta_path=("path to meta.json", "option", "m", str),
-    create_meta=("create meta.json, even if one exists in directory – if "
-                 "existing meta is found, entries are shown as defaults in "
-                 "the command line prompt", "flag", "c", bool),
-    force=("force overwriting of existing model directory in output directory",
-           "flag", "f", bool))
-def package(input_dir, output_dir, meta_path=None, create_meta=False,
-            force=False):
-    """
-    Generate Python package for model data, including meta and required
-    installation files. A new directory will be created in the specified
-    output directory, and model data will be copied over.
-    """
-    input_path = util.ensure_path(input_dir)
-    output_path = util.ensure_path(output_dir)
-    meta_path = util.ensure_path(meta_path)
-    if not input_path or not input_path.exists():
-        prints(input_path, title=Messages.M008, exits=1)
-    if not output_path or not output_path.exists():
-        prints(output_path, title=Messages.M040, exits=1)
-    if meta_path and not meta_path.exists():
-        prints(meta_path, title=Messages.M020, exits=1)
-
-    meta_path = meta_path or input_path / 'meta.json'
-    if meta_path.is_file():
-        meta = read_json(meta_path)
-        if not create_meta:  # only print this if user doesn't want to overwrite
-            prints(meta_path, title=Messages.M041)
-        else:
-            meta = generate_meta(input_dir, meta)
-    meta = validate_meta(meta, ['lang', 'name', 'version'])
-    model_name = meta['lang'] + '_' + meta['name']
-    model_name_v = model_name + '-' + meta['version']
-    main_path = output_path / model_name_v
-    package_path = main_path / model_name
-    bin_path = main_path / 'bin'
-    include_path = main_path / 'include'
-    orig_nc_path = Path(__file__).parent.parent
-    nc_path = package_path / 'neuralcoref'
-
-    create_dirs(package_path, force)
-    create_dirs(bin_path, force)
-    create_dirs(nc_path, force)
-
-    shutil.copytree(path2str(input_path),
-                    path2str(package_path / model_name_v))
-
-    orig_include_path = path2str(Path(__file__).parent / 'include')
-    shutil.copytree(path2str(orig_include_path),
-                    path2str(include_path))
-
-    nc1_path = path2str(orig_nc_path / 'neuralcoref.pyx')
-    nc2_path = path2str(orig_nc_path / 'neuralcoref.pxd')
-    shutil.copyfile(path2str(nc1_path),
-                    path2str(nc_path / 'neuralcoref.pyx'))
-    shutil.copyfile(path2str(nc2_path),
-                    path2str(nc_path / 'neuralcoref.pxd'))
-    create_file(nc_path / '__init__.py', TEMPLATE_INIT_NC)
-    create_file(nc_path / '__init__.pxd', TEMPLATE_INIT_PXD)
-
-    orig_bin_path = path2str(Path(__file__).parent.parent.parent / 'bin' / 'cythonize.py')
-    shutil.copyfile(path2str(orig_bin_path),
-                    path2str(bin_path / 'cythonize.py'))
-
-    create_file(main_path / 'meta.json', json_dumps(meta))
-    create_file(main_path / 'setup.py', TEMPLATE_SETUP)
-    create_file(main_path / 'MANIFEST.in', TEMPLATE_MANIFEST)
-    create_file(package_path / '__init__.py', TEMPLATE_INIT.format(model_name))
-    create_file(package_path / '__init__.pxd', TEMPLATE_INIT_PXD)
-    prints(main_path, Messages.M043,
-           title=Messages.M042.format(name=model_name_v))
-
-
-def create_dirs(package_path, force):
-    if package_path.exists():
-        if force:
-            shutil.rmtree(path2str(package_path))
-        else:
-            prints(package_path, Messages.M045, title=Messages.M044, exits=1)
-    Path.mkdir(package_path, parents=True)
-
-
-def create_file(file_path, contents):
-    file_path.touch()
-    file_path.open('w', encoding='utf-8').write(contents)
-
-
-def generate_meta(model_path, existing_meta):
-    meta = existing_meta or {}
-    settings = [('lang', 'Model language', meta.get('lang', 'en')),
-                ('name', 'Model name', meta.get('name', 'model')),
-                ('version', 'Model version', meta.get('version', '0.0.0')),
-                ('spacy_version', 'Required spaCy version',
-                 '>=%s,<3.0.0' % about.__version__),
-                ('description', 'Model description',
-                  meta.get('description', False)),
-                ('author', 'Author', meta.get('author', False)),
-                ('email', 'Author email', meta.get('email', False)),
-                ('url', 'Author website', meta.get('url', False)),
-                ('license', 'License', meta.get('license', 'CC BY-SA 3.0'))]
-    nlp = util.load_model_from_path(Path(model_path))
-    meta['pipeline'] = nlp.pipe_names
-    meta['vectors'] = {'width': nlp.vocab.vectors_length,
-                       'vectors': len(nlp.vocab.vectors),
-                       'keys': nlp.vocab.vectors.n_keys}
-    prints(Messages.M047, title=Messages.M046)
-    for setting, desc, default in settings:
-        response = util.get_raw_input(desc, default)
-        meta[setting] = default if response == '' and default else response
-    if about.__title__ != 'spacy':
-        meta['parent_package'] = about.__title__
-    return meta
-
-
-def validate_meta(meta, keys):
-    for key in keys:
-        if key not in meta or meta[key] == '':
-            prints(Messages.M049, title=Messages.M048.format(key=key), exits=1)
-    return meta
-
-
-TEMPLATE_SETUP = """
-#!/usr/bin/env python
-from __future__ import print_function
-import io
-import os
-from os import path, walk
-import json
-import sys
-import contextlib
-import subprocess
-from shutil import copy
-from distutils.sysconfig import get_python_inc
-from distutils import ccompiler, msvccompiler
-from setuptools import Extension, setup, find_packages
-
-PACKAGES = find_packages()
-
-PACKAGE_DATA = {'': ['*.pyx', '*.pxd']}
-
-def load_meta(fp):
-    with io.open(fp, encoding='utf8') as f:
-        return json.load(f)
-
-
-def list_files(data_dir):
-    output = []
-    for root, _, filenames in walk(data_dir):
-        for filename in filenames:
-            if not filename.startswith('.'):
-                output.append(path.join(root, filename))
-    output = [path.relpath(p, path.dirname(data_dir)) for p in output]
-    output.append('meta.json')
-    return output
-
-
-def list_requirements(meta):
-    parent_package = meta.get('parent_package', 'spacy')
-    requirements = [parent_package + ">=" + meta['spacy_version']]
-    if 'setup_requires' in meta:
-        requirements += meta['setup_requires']
-    return requirements
-
-@contextlib.contextmanager
-def chdir(new_dir):
-    old_dir = os.getcwd()
-    try:
-        os.chdir(new_dir)
-        sys.path.insert(0, new_dir)
-        yield
-    finally:
-        del sys.path[0]
-        os.chdir(old_dir)
-
-
-def generate_cython(root, source):
-    print('Cythonizing sources in', source)
-    p = subprocess.call([sys.executable,
-                         os.path.join(root, 'bin', 'cythonize.py'),
-                         source], env=os.environ)
-    if p != 0:
-        raise RuntimeError('Running cythonize failed')
-
-
-def is_source_release(model_path):
-    return os.path.exists(os.path.join(model_path, 'neuralcoref/neuralcoref.cpp'))
-
-
-def setup_package():
-    root = path.abspath(path.dirname(__file__))
-
-    with chdir(root):
-        meta_path = path.join(root, 'meta.json')
-        meta = load_meta(meta_path)
-        model_name = str(meta['lang'] + '_' + meta['name'])
-        model_dir = path.join(model_name, model_name + '-' + meta['version'])
-
-        include_dirs = [
-            get_python_inc(plat_specific=True),
-            os.path.join(root, 'include')]
-
-        if (ccompiler.new_compiler().compiler_type == 'msvc'
-            and msvccompiler.get_build_version() == 9):
-            include_dirs.append(os.path.join(root, 'include', 'msvc9'))
-
-        ext_modules = []
-        mod_name = model_name + '.neuralcoref.neuralcoref'
-        mod_path = mod_name.replace('.', '/') + '.cpp'
-        extra_link_args = []
-        # ???
-        # Imported from patch from @mikepb
-        # See Issue #267. Running blind here...
-        if sys.platform == 'darwin':
-            dylib_path = ['..' for _ in range(mod_name.count('.'))]
-            dylib_path = '/'.join(dylib_path)
-            dylib_path = '@loader_path/%s/neuralcoref/platform/darwin/lib' % dylib_path
-            extra_link_args.append('-Wl,-rpath,%s' % dylib_path)
-        ext_modules.append(
-            Extension(mod_name, [mod_path],
-                language='c++', include_dirs=include_dirs,
-                extra_link_args=extra_link_args))
-
-        if not is_source_release(model_name):
-            generate_cython(root, model_name)
-
-        copy(meta_path, path.join(model_name))
-        copy(meta_path, model_dir)
-        package_data = PACKAGE_DATA
-        package_data[model_name] = list_files(model_dir)
-        setup(
-            name=model_name,
-            description=meta['description'],
-            author=meta['author'],
-            author_email=meta['email'],
-            url=meta['url'],
-            version=meta['version'],
-            license=meta['license'],
-            ext_modules=ext_modules,
-            packages=PACKAGES,
-            package_data=package_data,
-            install_requires=list_requirements(meta),
-            zip_safe=False,
-        )
-
-
-if __name__ == '__main__':
-    setup_package()
-""".strip()
-
-
-TEMPLATE_MANIFEST = """
-include meta.json
-recursive-include include *.h
-recursive-include bin *.py
-""".strip()
-
-
-TEMPLATE_INIT = """
-# coding: utf8
-from __future__ import unicode_literals
-
-from pathlib import Path
-from spacy.util import load_model_from_init_py, get_model_meta
-from {}.neuralcoref import NeuralCoref
-
-__version__ = get_model_meta(Path(__file__).parent)['version']
-
-
-def load(**overrides):
-    disable = overrides.get('disable', [])
-    if 'neuralcoref' in disable:
-        nlp = load_model_from_init_py(__file__, **overrides)
-    else:
-        overrides['disable'] = disable + ['neuralcoref']
-        nlp = load_model_from_init_py(__file__, **overrides)
-        coref = NeuralCoref(nlp.vocab)
-        coref.from_disk(nlp.path / 'neuralcoref')
-        nlp.add_pipe(coref, name='neuralcoref')
-    return nlp
-""".strip()
-
-TEMPLATE_INIT_NC = """
-from .neuralcoref import NeuralCoref
-""".strip()
-
-TEMPLATE_INIT_PXD = """
-""".strip()
-
diff --git a/neuralcoref/file_utils.py b/neuralcoref/file_utils.py
index da00ca1..ef027f9 100644
--- a/neuralcoref/file_utils.py
+++ b/neuralcoref/file_utils.py
@@ -31,6 +31,9 @@
 except (AttributeError, ImportError):
     NEURALCOREF_CACHE = os.getenv('NEURALCOREF_CACHE', os.path.join(os.path.expanduser("~"), '.neuralcoref_cache'))
 
+NEURALCOREF_MODEL_URL = "https://s3.amazonaws.com/models.huggingface.co/neuralcoref/neuralcoref.tar.gz"
+NEURALCOREF_MODEL_PATH = os.path.join(str(NEURALCOREF_CACHE), "neuralcoref")
+
 logger = logging.getLogger(__name__)  # pylint: disable=invalid-name
 
 
diff --git a/neuralcoref/neuralcoref.pyx b/neuralcoref/neuralcoref.pyx
index d3c5b31..30d6789 100644
--- a/neuralcoref/neuralcoref.pyx
+++ b/neuralcoref/neuralcoref.pyx
@@ -18,7 +18,6 @@ cimport cython
 from cpython cimport array
 import array
 from libc.stdint cimport uint16_t, uint32_t, uint64_t, uintptr_t, int32_t
-import cytoolz
 
 import numpy
 from cymem.cymem cimport Pool
@@ -44,6 +43,8 @@ from thinc.v2v import Model, ReLu, Affine
 from thinc.api import chain, clone
 # from thinc.neural.util import get_array_module
 
+from .file_utils import NEURALCOREF_MODEL_PATH
+
 ##############################
 ##### A BUNCH OF SIZES #######
 
@@ -542,6 +543,9 @@ cdef class NeuralCoref(object):
             Token.set_extension('coref_clusters', getter=self.token_clusters)
             Token.set_extension('coref_scores', getter=self.token_scores)
 
+        # Load from disk
+        self.from_disk(NEURALCOREF_MODEL_PATH)
+
     def __reduce__(self):
         return (NeuralCoref, (self.vocab, self.model), None, None)
 
@@ -595,7 +599,7 @@ cdef class NeuralCoref(object):
             conv_dict = self.cfg.get('conv_dict', None)
         if blacklist is None:
             blacklist = self.cfg.get('blacklist', True)
-        for docs in cytoolz.partition_all(batch_size, stream):
+        for docs in util.minibatch(stream, size=batch_size):
             docs = list(docs)
             annotations = self.predict(docs, greedyness=greedyness, max_dist=max_dist,
                                     max_dist_match=max_dist_match, blacklist=blacklist)
diff --git a/neuralcoref/tests/__init__.py b/neuralcoref/tests/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/requirements.txt b/requirements.txt
index ed70eb4..e571c6f 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,3 +1,5 @@
 spacy
 cython
-numpy
\ No newline at end of file
+numpy
+boto3
+requests
\ No newline at end of file
diff --git a/setup.py b/setup.py
index b845fe0..e52a46f 100644
--- a/setup.py
+++ b/setup.py
@@ -1,16 +1,31 @@
 #!/usr/bin/env python
 from __future__ import print_function
+import io
 import os
 import subprocess
 import sys
 import contextlib
-import textwrap
-import pkg_resources
 from distutils.command.build_ext import build_ext
 from distutils.sysconfig import get_python_inc
+import distutils.util
 from distutils import ccompiler, msvccompiler
 from setuptools import Extension, setup, find_packages
 
+def is_new_osx():
+    """Check whether we're on OSX >= 10.10"""
+    name = distutils.util.get_platform()
+    if sys.platform != "darwin":
+        return False
+    elif name.startswith("macosx-10"):
+        minor_version = int(name.split("-")[1].split(".")[1])
+        if minor_version >= 7:
+            return True
+        else:
+            return False
+    else:
+        return False
+
+
 PACKAGE_DATA = {'': ['*.pyx', '*.pxd'],
                 '': ['*.h'],}
 
@@ -20,21 +35,79 @@
 
 MOD_NAMES = ['neuralcoref.neuralcoref']
 
-def is_installed(requirement):
-    try:
-        pkg_resources.require(requirement)
-    except pkg_resources.ResolutionError:
-        return False
-    else:
-        return True
 
-if not is_installed('numpy>=1.11.0') or not is_installed('spacy>=2.0.4'):
-    print(textwrap.dedent("""
-            Error: requirements needs to be installed first.
-            You can install them via:
-            $ pip install -r requirements.txt
-            """), file=sys.stderr)
-    exit(1)
+
+COMPILE_OPTIONS = {
+    "msvc": ["/Ox", "/EHsc"],
+    "mingw32": ["-O2", "-Wno-strict-prototypes", "-Wno-unused-function"],
+    "other": ["-O2", "-Wno-strict-prototypes", "-Wno-unused-function"],
+}
+
+
+LINK_OPTIONS = {"msvc": [], "mingw32": [], "other": []}
+
+
+if is_new_osx():
+    # On Mac, use libc++ because Apple deprecated use of
+    # libstdc
+    COMPILE_OPTIONS["other"].append("-stdlib=libc++")
+    LINK_OPTIONS["other"].append("-lc++")
+    # g++ (used by unix compiler on mac) links to libstdc++ as a default lib.
+    # See: https://stackoverflow.com/questions/1653047/avoid-linking-to-libstdc
+    LINK_OPTIONS["other"].append("-nodefaultlibs")
+
+
+USE_OPENMP_DEFAULT = "0" if sys.platform != "darwin" else None
+if os.environ.get("USE_OPENMP", USE_OPENMP_DEFAULT) == "1":
+    if sys.platform == "darwin":
+        COMPILE_OPTIONS["other"].append("-fopenmp")
+        LINK_OPTIONS["other"].append("-fopenmp")
+        PACKAGE_DATA["spacy.platform.darwin.lib"] = ["*.dylib"]
+        PACKAGES.append("spacy.platform.darwin.lib")
+
+    elif sys.platform == "win32":
+        COMPILE_OPTIONS["msvc"].append("/openmp")
+
+    else:
+        COMPILE_OPTIONS["other"].append("-fopenmp")
+        LINK_OPTIONS["other"].append("-fopenmp")
+
+
+# By subclassing build_extensions we have the actual compiler that will be used which is really known only after finalize_options
+# http://stackoverflow.com/questions/724664/python-distutils-how-to-get-a-compiler-that-is-going-to-be-used
+class build_ext_options:
+    def build_options(self):
+        for e in self.extensions:
+            e.extra_compile_args += COMPILE_OPTIONS.get(
+                self.compiler.compiler_type, COMPILE_OPTIONS["other"]
+            )
+        for e in self.extensions:
+            e.extra_link_args += LINK_OPTIONS.get(
+                self.compiler.compiler_type, LINK_OPTIONS["other"]
+            )
+
+
+class build_ext_subclass(build_ext, build_ext_options):
+    def build_extensions(self):
+        build_ext_options.build_options(self)
+        build_ext.build_extensions(self)
+
+
+# def is_installed(requirement):
+#     try:
+#         pkg_resources.require(requirement)
+#     except pkg_resources.ResolutionError:
+#         return False
+#     else:
+#         return True
+
+# if not is_installed('numpy>=1.11.0') or not is_installed('spacy>=2.1.0'):
+#     print(textwrap.dedent("""
+#             Error: requirements needs to be installed first.
+#             You can install them via:
+#             $ pip install -r requirements.txt
+#             """), file=sys.stderr)
+#     exit(1)
 
 @contextlib.contextmanager
 def chdir(new_dir):
@@ -70,11 +143,11 @@ def setup_package():
 
         include_dirs = [
             get_python_inc(plat_specific=True),
-            os.path.join(root, 'neuralcoref', 'cli', 'include')]
+            os.path.join(root, 'include')]
 
         if (ccompiler.new_compiler().compiler_type == 'msvc'
             and msvccompiler.get_build_version() == 9):
-            include_dirs.append(os.path.join(root, 'neuralcoref', 'cli', 'include', 'msvc9'))
+            include_dirs.append(os.path.join(root, 'include', 'msvc9'))
 
         ext_modules = []
         for mod_name in MOD_NAMES:
@@ -94,30 +167,41 @@ def setup_package():
                     extra_link_args=extra_link_args))
 
         setup(name='neuralcoref',
-            version='3.1',
+            version='4.0',
             description="Coreference Resolution in spaCy with Neural Networks",
             url='https://github.com/huggingface/neuralcoref',
             author='Thomas Wolf',
             author_email='thomwolf@gmail.com',
             ext_modules=ext_modules,
-            include_dirs=[numpy.get_include()],
             classifiers=[
                 'Development Status :: 3 - Alpha',
                 'Environment :: Console',
                 'Intended Audience :: Developers',
-                'Programming Language :: Python',
-                'Programming Language :: Python :: 2.7',
-                'Programming Language :: Python :: 3.3',
-                'Programming Language :: Python :: 3.4',
-                'Programming Language :: Python :: 3.6'
+                "Intended Audience :: Science/Research",
+                "License :: OSI Approved :: MIT License",
+                "Operating System :: POSIX :: Linux",
+                "Operating System :: MacOS :: MacOS X",
+                "Operating System :: Microsoft :: Windows",
+                "Programming Language :: Cython",
+                "Programming Language :: Python :: 2",
+                "Programming Language :: Python :: 2.7",
+                "Programming Language :: Python :: 3",
+                "Programming Language :: Python :: 3.4",
+                "Programming Language :: Python :: 3.5",
+                "Programming Language :: Python :: 3.6",
+                "Programming Language :: Python :: 3.7",
+                "Topic :: Scientific/Engineering",
             ],
-            install_requires=['numpy', 'spacy>=2.1.0', 'cytoolz'],
+            install_requires=['numpy>=1.15.0', 'spacy>=2.1.0'],
+            setup_requires=["wheel"],
+            python_requires=">=2.7,!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.*",
             packages=PACKAGES,
             package_data=PACKAGE_DATA,
             keywords='NLP chatbots coreference resolution',
             license='MIT',
             zip_safe=False,
-            platforms='any')
+            platforms='any',
+            cmdclass={"build_ext": build_ext_subclass})
 
 if __name__ == '__main__':
     setup_package()