checkout to version v0.1.1

doc/release-note.md

+## v0.1.1
+
+### Distributed
+
+- Broadcast data-parallel parameters before training.
+
+### Megatron adaption
+
+- Initialize `FMoELinear` parameters using different seed in model parallel even using the same random seed in megatron.
+- Use proper comm for mp and dp.
+
+### Transformer-XL example
+
+- Improve scripts.
+
+### Misc
+
+- Logo and slack workspace link.
+- Document in Chinese.
+- Figures to explain how FastMoE works.
+
 ## v0.1.0
 
 ### Functions

setup.py

@@ -15,9 +15,9 @@ if os.environ.get('USE_NCCL', '0') == '1':
 
 if __name__ == '__main__':
     setuptools.setup(
-        name='fmoe',
-        version='0.1.0',
-        description='An efficient Mixture-of-Experts impl. for PyTorch',
+        name='fastmoe',
+        version='0.1.1',
+        description='An efficient Mixture-of-Experts system for PyTorch',
         author='Jiaao He, Jiezhong Qiu and Aohan Zeng',
         author_email='hja20@mails.tsinghua.edu.cn',
         license='Apache-2',