AI/PyTorch Partner Engineer - Meta AI (Facebook AI)
Principal Software Engineer - Audere
Software Architect - X10 Wireless
Dev/PM - Microsoft
- Seattle, WA USA
- https://www.linkedin.com/in/lessw2020
Popular repositories Loading
-
Ranger-Deep-Learning-Optimizer
Ranger-Deep-Learning-Optimizer PublicRanger - a synergistic optimizer using RAdam (Rectified Adam), Gradient Centralization and LookAhead in one codebase
-
Best-Deep-Learning-Optimizers
Best-Deep-Learning-Optimizers PublicCollection of the latest, greatest, deep learning optimizers (for Pytorch) - CNN, NLP suitable
-
res2net-plus
res2net-plus PublicRes2Net architecture with improved stem and Mish activation function
-
Ranger-Mish-ImageWoof-5
Ranger-Mish-ImageWoof-5 PublicRepo to build on / reproduce the record breaking Ranger-Mish-SelfAttention setup on FastAI ImageWoof dataset 5 epochs
1,861 contributions in the last year
Day of Week | March Mar | April Apr | May May | June Jun | July Jul | August Aug | September Sep | October Oct | November Nov | December Dec | January Jan | February Feb | |||||||||||||||||||||||||||||||||||||||||
Sunday Sun | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Monday Mon | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Tuesday Tue | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Wednesday Wed | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Thursday Thu | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Friday Fri | |||||||||||||||||||||||||||||||||||||||||||||||||||||
Saturday Sat |
Less
No contributions.
Low contributions.
Medium-low contributions.
Medium-high contributions.
High contributions.
More
Contribution activity
March 2025
Created 2 commits in 1 repository
Reviewed 7 pull requests in 1 repository
pytorch/torchtitan
7 pull requests
-
[rocm] add amd gpu peak flops info
This contribution was made on Mar 4
-
fix MI250X peak FLOPS
This contribution was made on Mar 4
-
update path of
train.py
totorchtitan/train.py
This contribution was made on Mar 4 -
[Checkpointing] Force to save a checkpoint at step 1 if checkpoint is enabled
This contribution was made on Mar 4
-
[Checkpointing] Change the default keep_last_k to 10
This contribution was made on Mar 4
-
Add FSSpec to the dependency
This contribution was made on Mar 4
-
[DeepSeek] Enable checkpoint load from HF
This contribution was made on Mar 1
Created an issue in pytorch/torchtitan that received 2 comments
[Checkpointing] Set reasonable default for keep_latest_k checkpoints and avoid overflowing user drives...currently set to 'infinite' checkpoints
Bug description The current setting for keep_latest_k = 0, which means 'infinite' checkpoints will be saved. This is a poor default and endangers t…
2
comments
Opened 6 other issues in 1 repository
pytorch/torchtitan
5
open
1
closed
-
[Feature] Add Multi-Token Prediction module
This contribution was made on Mar 5
-
[Feature] add preflight NCCL and GEMM check to multinode slurm script
This contribution was made on Mar 3
-
[Feature] expose Torch Nan checker as configurable option in toml for those training at scale
This contribution was made on Mar 3
-
[Checkpointing][feature] add option for pre-flight checkpoint verification
This contribution was made on Mar 3
-
[Checkpointing] fails out if checkpoint folder does not exist when using keep_latest_k
This contribution was made on Mar 2
-
[Checkpointing] Using keep_latest_k setting results in failure when using external mounted drive
This contribution was made on Mar 1
11
contributions
in private repositories
Mar 2 – Mar 4