noodlepotato


























  1. If you're looking for an easy way to run LLM inference with this model locally, you can try

  2. what about fine-tuning this Tinyllama chat in txtai? probably using unknown domains dataset

  3. I was using Mistral instruct 7b @ Q4 on my iPhone 15 Pro during a flight (no internet) recently. The thing was almost useless and schizophrenic...

  4. May I know how are you running this? Interested to try on my 15 too

  5. Hey, try to do NLP + RL instead of pure RL because there is a clear shift where people are moving away from RL and trying to integrate RL stuff in NLP. I have a lot of people in my institute ( one of the top in RL) doing that. I heard that people in deep mind are also moving away working more on foundation models.

  6. Any examples/links/source for this? Of course aside from the RLHF PPO stuff

  7. My advice: use your pungbeak! He is very good against mons that prevent damage (molly, triana, water cookie) by just attacking that target and preventa revive of the initially attacked one. Pung Tiana Gale Deborah is really strong against slow and tanky comps and revivers.

  8. Oh nice let me try that. I never use my pungbeak so it doesn't have any runes because I thought my cleavers are enough. Thanks for the suggestion!

  9. AO: Man you have Nephthys Tiana core I'm so jealous lol. Almost all of your AO will work up to P3. Depends in your rune of course.

  10. I'm very lucky on Nep but I mostly use it for TOA lol and got Tiana from the nat5 event.

  11. Hi everyone! I'm currently a 28-year-old data clerk turned data scientist with a total of 5 years of experience under my belt. I'm also in my first year of pursuing a PhD in Data Science. My work primarily revolves around LLM and CV, and most of the time, we only utilize pretrained models, make some small adjustments, and deploy them into the cloud. However, I don't want this to be my long-term career path because I have a strong interest in other fields, particularly reinforcement learning and time-series analysis.

  12. Hi everyone! I'm currently a 28-year-old data clerk turned data scientist with a total of 5 years of experience under my belt. I'm also in my first year of pursuing a PhD in Data Science. My work primarily revolves around LLM and CV, and most of the time, we only utilize pretrained models, make some small adjustments, and deploy them into the cloud. However, I don't want this to be my long-term career path because I have a strong interest in other fields, particularly reinforcement learning and time-series analysis.

  13. Causal inference is a really big field with tons of techniques with most of them (with the exception of RCTs) suffering from the same pitfall, they’re very sensitive to the technique you are using. Different techniques will you give you different results. Oftentimes what we do in my role is basically look at a bunch of techniques and if they all show the same directional result we call it a win. I’ll talk about some of the techniques I use in my own role.

  14. Causal machine learning is extremely active. Susan Athey and her coauthors are publishing banger papers every other week.

  15. Do you know where should I start on this one? Tired of Gen AI and typical xgboost models.

  16. You meant Flex? Would that be a valid choice also?

  17. Probably! Though I wouldn't count on it. I linked another forum post about the WiFi drivers.

  18. Thank you for the research! Let me look at it. Maybe I can crosspost to Linux Mint in advance.

  19. Based on one commenter, in the other subreddit i posted, samsung laptop are experiencing wifi issues

  20. As it should be. Also with pvp titles e.g Archeage and Guild wars 2, probably more prestige than any end games set.

  21. Quite weird na nowadays na entering IT is like a choice of breakfast nalang. I think most of people na gusto pumasok have no idea what it takes para mag succeed sa tech field. You can easily get exposed if di mo alam alam ginagawa mo. If you finished a Udemy course, it doesn’t mean you’re good enough na to find a high paying job. Tech field is very meritocratic unlike courses like engineering where “pa exp ka muna” is the most thing you would hear. Yes mataas sweldo but you can easily get burned out, like sunog.

  22. You can play anything you can play on a Windows PC.

  23. Thanks. I thought ROG Ally has a “handheld” version of windows PC hence some limitations.

  24. Nah but Im on the premium pro annual. Queue happened back in Christmas and New Year but usual day is rarely, I can say I never exp queue. Where you based?

  25. been using GeForceNOW SG since last year. I have 100-300 mbps internet, but the lag at the evening is bad

  26. I used txtai for information retrieval few days ago. So far it’s intuitive to use and has like 30+ examples, very easy to setup for local and do some POC but I’m not sure for production.

  27. Ironically, I got less work load when I managed to go up in the corporate ladder and I’m just 27. Probably a different situation talaga sa DS field. Mag kukunwari lang ako ng seryoso sa scrum tas okay na. My senior is also having lesser work load than me. E.g mostly meetings na again, magkukunwari lang kami seryoso haha (and because automated na halos lahat ng task namin)

  28. Update: It's been a little tricky but Polars is quite amazing. Getting used to the new API takes a lot of effort at first because it's really different from Pandas. There are still some features that are lacking, quarterly resampling doesn't natively exist but it's a very basic use case for any financial series. The speed is amazing though and it's great that the core devs are focused on it.

  29. Same here. I used Pandas for financial stuff but I used polars when dealing with less "tabular" e.g text data with some features. Pandas is really superior for me when it comes to tabular data because of the ecosystem like sktime, feature-engine and pyod.

Leave a Reply

Your email address will not be published. Required fields are marked *

Author: admin