Comments (2)
Hi @jzhang38 sorry for being slow. I was on vacation.
It would be nice to have a public table that's continously updated or something like that. I'll share this internal doc that @vadam5 was fleshing out. https://docs.google.com/spreadsheets/d/1H6Ev3FY0d4VLUM6G4igVClIS296pJ7MrSpajZvmVCOQ/edit?usp=sharing
There's something wrong with the 13B runs (the nodes we were using failed before we sorted it out).
We don't have regular access to H100s (yet) so can't really do anything there besides the numbers I've already shared.
We're still honestly mostly a TPU shop, except for smaller, mostly single node runs.
from levanter.
Thanks for sharing!
from levanter.
Related Issues (20)
- Add Mixtral support HOT 3
- Recommended citation for Levanter? HOT 4
- Dependency Issues with Scipy and Pyaml-Include HOT 4
- Add TPU Metrics to Weights and Biases Logging HOT 3
- Redo eval loss loop to loop through all domains at once, support tags HOT 2
- add bpb metric
- LoRA layers via dispatch HOT 4
- Clean up the story for overriding hf configs (e.g. for scan_layers)
- wire up llm-eval-harness
- Llama 3 HOT 20
- Possible Ray OOM when procesing long documents HOT 5
- Scanning layers seems not to allow the mixing of precisions in mp anymore
- Tensorboard FileExistsError on SLURM when training multi-node with 1 process per GPU HOT 5
- Add requirements.txt (poetry toml) for tests
- Double BOS token on Mistral HOT 2
- Training job hangs after saving checkpoint due to curl timeout error HOT 1
- Silent failure at the data tokenization pipeline due to duplicate file names HOT 1
- Hang when training LLama7B on TPU v5e-16 HOT 9
- Issues with Splash Attention HOT 9
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from levanter.