Jeff

Full Stack Software Developer @NorthropGrumman.
Hobbies: Physics, AI,ML,DL & Robotics & #IoT.

[Jeff @Jeff_Hamlin]

@stephenasmith Didn't you pick them Cowturds to win? As much as I look forward to these, it seems somewhat hypocritical to me to celebrate the choke, but you predicted a win.

[Stephen A Smith @stephenasmith]

How ‘bout them Cowboys!!!

[Jeff @Jeff_Hamlin]

@Espngreeny you are so right about computers being better rational decision makers than humans, but only in specific situations.

[Jeff @Jeff_Hamlin]

@keyshawn He's worse than when he was at Oklahoma

[Thomas Wolf @Thom_Wolf]

Over the past weeks the H4 team has been busy pushing the Zephyr 7B model to new heights 🗻

The new version is now topping all 7b models on chat evals and even 10x larger models 🤯🔥

Here are the intuitions on it

1/ Start with the strongest pretrained model you can find: Mistral 7B is amazing, by far the strongest 7B pretrained model. Props to the Mistral AI team 😍

2/ Scale human-preference annotations: several studies have show how for many tasks GPT4 is on-par with the average human annotators while making scalable annotations as easy as an API call: the H4 team started from the largest and most diverse public GPT4 preference annotation dataset: UltraFeedback 🤖🦾

3/ Drop Reinforcement Learning in favor of DPO (Direct Preference Optimization): while using LLM in RL is definitely much easier compared to the struggles of getting deep-RL to work from scratch, DPO totally remove RL from the preference annotation training and directly optimize the preference model in a much more stable training procedure in the H4 team's experiments

4/ Don't be scared of overfitting on preference dataset: This is maybe the most counterintuitive results of the work. While the train/test loss of DPO training shows signs of overfitting on the feedback dataset after just one epoch, training further still show significant improvements on downstream tasks even up to 3 epochs without signs of performances regression. Would be interesting to dive even further in this surprising behavior

5/ Share everything openly 😁 the recipes, code, model and dataset will be available at https://t.co/qkKhS8eqvt
In the meantime the paper is a great starting point: https://t.co/qzZ0tb8Aq1

[Jeff @Jeff_Hamlin]

@shahedC The tech articles make me think it's highly popular, but I don't have it and so far, don't miss it

[Jeff @Jeff_Hamlin]

@jasrifootball I don't understand why this is Russell fault if he's been running all game but now he's supposed to see a linebacker coming up fast from behind?

[Jeff @Jeff_Hamlin]

@shahedC It's a super nice computer but overpowered for your son - way overpowering 😇😁

[Jeff @Jeff_Hamlin]

@MichelleObama @BarackObama

[Jeff @Jeff_Hamlin]

@EmmanuelAcho Ignore this guy, he likes to stir you up

[Jeff @Jeff_Hamlin]

@TheRevAl Reverend, on @JoeNBC they seemed confused about why minorities aren't jumping to support Biden. He was more interested in his infrastructure bill than protecting our voting rights or changing some aspects of policing like qualified immunity. SCOTUS did more. In Al.

[Jeff @Jeff_Hamlin]

@ColinCowherd you don't like @DangeRussWilson you have framed your narrative to support that conclusion. You will never say a bad word about Sean Payton so your bias kills any argument you might make. It's the same bias about Justin Fields.

[Jeff @Jeff_Hamlin]

@Darth__Rose @PatMcAfeeShow @OfficialAJHawk McAfee blamed Jameis, what does a stupid punter know 🤔 😑 🙄
Jameis put the team in position to win, and a kicker missed ..

[Kirk Herbstreit @KirkHerbstreit]

FULL Corso 400th Headgear piece 🫶🏼

[Stephen A Smith @stephenasmith]

Haaaaaaa

[Jeff @Jeff_Hamlin]

@stephenasmith @mspears96 , @SHAQ Hello? Great Defense 🤣😂🤣😂

[Stephen A Smith @stephenasmith]

It is a Bird….it’s not a plane!!!! It’s the cardinals??? 🤔😭😭😭

Had to bring this back 😭