lewq

9 ms·

61.
by lewq 3y ago
Fine tuning is just more training -- so it's definitely possible to teach the model facts this way too. In practice we've found that it's a bit of a balancing act to teach the model the new knowledge without destroying existi
62.
How we got fine-tuning Mistral-7B to not suck(helixml.substack.com)
215 points by lewq 3y ago|47 comments
63.
Type error in AI? Can models be intelligent or are they the same thing as maps?(hodesdon.substack.com)
5 points by lewq 3y ago|4 comments
64.
by lewq 3y ago
Type error in AI? Can models be intelligent or are they the same kind of thing as maps? Interesting new blog -- https://hodesdon.substack.com/about https://hodesdon.substack.com/p/what-ai-is-and-is-not
65.
UI for fine tuning Mistral and SDXL, GPU mem/latency optimization(docs.helix.ml)
2 points by lewq 3y ago|1 comments
66.
by lewq 3y ago
100% bootstrapped new startup with source available on GitHub. It lets you fine tune Mistral-7B and SDXL with a nice UI. In particular, for the LLM fine tuning we implemented a dataprep pipeline that turns websites/pdfs/doc files
67.
by lewq 3y ago
Thank you! <3
68.
by lewq 3y ago
Some resources: Demo: https://www.youtube.com/watch?v=Ym4nPSzfer0 About Helix Helix is a generative AI platform that you can run on our cloud or deploy in your own data center or cloud account. It provides an easy-to-use in
69.
Show HN: Local fine tuning for Mistral and SDXL, GPU mem/latency optimization(docs.helix.ml)
36 points by lewq 3y ago|3 comments
70.
by lewq 4y ago
I thought this was cool because it's very different from the usual anime crap And you can see how smoothly varying the math produces these varying images in response to the prompt
71.
Understanding classifier-free guidance for style transfer(hodesdon.com)
29 points by lewq 4y ago|6 comments
72.
by lewq 4y ago
This is the kind of high quality technical content I come here for.
73.
by lewq 5y ago
Are you on an internal ML platform team? We'd love to know if self-service (for data scientists), multi-tenant model -> container image builds are a pain point for you. I was lucky enough to be involved in the creation of https:&#x
74.
Show HN: MLflow to KFServing image builds with Kaniko on K8s(chassis.ml)
3 points by lewq 5y ago|1 comments
75.
by lewq 5y ago
Thanks, would love to know how you get on, please join #boxkite-monitoring on MLOps.community slack to let us know how it works out for you :-)
76.
by lewq 5y ago
We built boxkite because lots of ML drift monitoring solutions are too complex and heavyweight: this is a pure python library exposing a pure prometheus endpoint, then we implemented KL divergence in one massive PromQL query :-)
77.
ML data and model drift in Prometheus and Grafana(boxkite.ml)
24 points by lewq 5y ago|5 comments
78.
by lewq 6y ago
1. Deploying models easily (data scientist doesn't need to grok docker/kube) 2. Monitoring models including data and model drift/statistical monitoring (data scientists don't need to grok prom/grafana) 3. Only once
79.
by lewq 6y ago
Dotmesh will remain open source and as the original author I'm happy to help maintain it. Are you using it or interested in using it?
80.
by lewq 6y ago
I'd love to, genuinely, but I don't think it will be possible. https://github.com/dotmesh-io/dotmesh is already open source though which is one key component of Dotscience.
81.
by lewq 6y ago
We priced too low at first, tried too high when we released our enterprise product, did some compares with other dev tooling in sales convos with early customers and then settled on something in the middle ($600/user/month) and st
82.
by lewq 6y ago
We had early revenue and healthy pipeline, we just ran out of that crucial currency: investor patience.
83.
by lewq 6y ago
Thank you
84.
by lewq 6y ago
Happy to talk - please email me - luke@dotscience.com
85.
by lewq 6y ago
Direct outreach on LinkedIn and conferences. Fintech & insurance tech was beachhead market, followed by finance.
86.
by lewq 6y ago
Btw https://mlops.community (which we started) will live on, Demetrios, Chris and Dan are kindly keeping it running ️ If you're interested in MLOps, come and join our weekly 9am PT / 5pm UK Wednesday online meetups - i
87.
by lewq 6y ago
CEO here. Happy to make introductions to the team for anyone who is hiring. We have valuable residual knowledge about building a product in the MLOps platform space and achieving product-market fit, and I hope that this might be helpful to
88.
Dotscience is shutting down(dotscience.com)
74 points by lewq 6y ago|45 comments
89.
by lewq 7y ago
Nice article!
90.
by lewq 7y ago
It's an environment for tracking runs of data engineering and ML training code in, so you can use tools like H2O and sklearn grid search within Dotscience. But we could also build on the run mechanism to automate kicking off a lot of r

More ›