💻 Research Interests in Machine Learning and Distributed System📚 Active learner🕹 Game lover🏀 Lakers fan
FrankLeeeee
Follow
FrankLeeeee
Follow
Popular repositories
-
OmniViewer Public
Powerful media viewer for Computer Vision engineers working on remote servers.
-
-
ColossalAI Public
Forked from hpcaitech/ColossalAI
Colossal-AI: A Unified Deep Learning System for Large-Scale Parallel Training
Python 1
-
ColossalAI-Examples Public
Forked from hpcaitech/ColossalAI-Examples
Examples of training models with hybrid parallelism using ColossalAI
-
Crawl annual report PDFs for Chinese companies listed in Shenzhen
Python 1
-
Solutions to Andrew Ng's Coursera course assignments on Machine Learning
MATLAB
489 contributions in the last year
Less
More
Activity overview
Contributed to
hpcaitech/ColossalAI,
hpcaitech/ColossalAI-Examples,
NUS-HPC-AI-Lab/oh-my-server
and 5 other
repositories
Contribution activity
March 2022
Created 32 commits in 5 repositories
Created 1 repository
Created a pull request in hpcaitech/ColossalAI that received 8 comments
Opened 23 other pull requests in 3 repositories
hpcaitech/ColossalAI
2
open
14
merged
2
closed
- refactored fp16 optimizer
- Merge a missing fix into main
- [doc] fixed broken badge link
- Merge develop into main
- fixed bug in activation checkpointing test
- [amp] refactored fp16 optimizer
- [unit test] Refactored test cases with component func
- [fp16] refactored grad scaler
- [fix] set criterion as optional in colossalai initialize
- refactored testing components
- refactored testing components
- [zero] add more shard strategy
- Refactored github action
- Refactored CI for new git workflow
- fixed typo in ShardParam
- added unit test for sharded optimizer
- remove deepspeed implementation and refactor for the reconstructed zero module
- added compatibility CI and options for release CI
hpcaitech/ColossalAI-Benchmark
2
merged
1
closed
NUS-HPC-AI-Lab/oh-my-server
2
merged
Reviewed 42 pull requests in 3 repositories
hpcaitech/ColossalAI
31 pull requests
- [zero] polish ShardedOptimV2 unittest
- Fix/format
- [zero] cuda memory usage tracer
- [zero] able to place params on cpu after zero init context
- [test] polish zero related unitest
- [zero] add test sharded optim with cpu adam
- [zero] move async memory to an individual directory
- Added Profiler Context to manage all profilers
- [zero] fix bert unit test
- added base profiler
- [profiler] add adaptive sampling to memory profiler
- fix format for dir-[parallel_3d]
- [zero] update sharded optim v2
- [test] pytest parametrize
- [zero] update zero context init with the updated test utils
- fixed strings in profiler outputs
- [zero] zero init context
- [profile] add profiler communication operations
- add sharded optim v3
- add badge and contributors
- [zero] add more shard strategy
- [zero] polish shard strategy
- [zero] cpu adam kernel
- [zero] a shard strategy in granularity of tensor
- add sharded adam
- Some pull request reviews not shown.
hpcaitech/ColossalAI-Benchmark
8 pull requests
hpcaitech/ColossalAI-Examples
3 pull requests
Created an issue in hpcaitech/ColossalAI that received 2 comments
[RFC] Provide Communication Profiler as a Context Manager
The current communication profiler works in a global scope as we need to initialize the profiler before training. Can we provide a context manager …
2
comments
Opened 2 other issues in 2 repositories
hpcaitech/ColossalAI-Benchmark
1
open
hpcaitech/ColossalAI
1
closed
9
contributions
in private repositories
Mar 1 – Mar 7