lpasselin

Joined 10 years ago

Karma
175
Submissions
101
Member since
Sep 25
[ my public key: https://keybase.io/lpasselin; my proof: https://keybase.io/lpasselin/sigs/YWIITO_iIUEVT5paUBS-NgUR_GZaLG167gL3Rh3Aagg ]

https://asselin.engineer

View on Hacker News

Recent submissions

Comment · 1 year ago
A lawyer can become a locksmith in a few months (weeks? days?). If the flip mentioned is really happening, isn't it temporary?
Comment · 1 year ago
Hey, I read the paper in detail and presented to colleagues during our reading group.

I still do not understand exactly where D1L comes from in LGan(D1L, T(A1(u)). Is D1L simply A1(u)?

I also find that mixing notation in figure 2 and 3 makes it tricky.

Would have loved to have more insights from the results in the tables.

And more results from inversion, on more than Enron dataset. Since that is one end goals, even if reusing another method.

Thank you for the paper, very interesting!

Comment · 2 years ago
Do the captured elements move a lot during this snapshot? since it will take months? Is the difference significant?
Comment · 2 years ago
This comes from the same group as the EfficientViT model. A few months ago, their EfficientViT model was the only modern and small ViT style model I could find that had raw pytorch code available. No dependencies to the shitty framework and libraries that other ViT are using.
Comment · 2 years ago
Custom and _competent_ AI tutors will be a game changer for education.
Comment · 2 years ago
Meanwhile, I had a hard time last week getting a machine with 8 gpus from azure.
Comment · 2 years ago
I have a 200$ mini machine and would like to upgrade. Can DIY anything. What would you suggest for a maximum budget of 1500$?
Comment · 2 years ago
The mamba paper shows significant improvements in all model sizes, up to 1b, the largest one tested.

Are there any reason why it wouldn't scale to 7b or more? Have they tried it?

Comment · 2 years ago
Most of the leaderboard has much lower sequence length.
Comment · 3 years ago
Last time I tried, it did not have a good android app working.
Comment · 3 years ago
What are these websites??
Comment · 3 years ago
I actually tried this last year, before OpenAI released their cheaper embeddings v2 in december. From my experiments, when compared to Bert embeddings (or recent variation of the model) the OpenAI embeddings are miles ahead when doing similarity search.
Comment · 4 years ago
Comment · 4 years ago
There are attachments that insert directly into the battery hole which are connected to a power supply.
Comment · 4 years ago
Are they releasing a new iPad pro soon? Buyers guide says It's been 495 days.

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton