Lena

Trying AI tools in real workflows and sharing what actually helps. Notes on creating, building, and working a little smarter.

[Lena @_LenaHart]

AI won’t replace everyone.
But people who know how to use AI will have an advantage over those who refuse to learn it.
Learn the tools. Build the skill. Stay ahead.

[Iris Quinn @Iriss_Quinn]

I’d love to see the landing page, deck and social posts made from one conversation. That’s the workflow I want to watch.

[Iris Quinn @Iriss_Quinn]

A model checkpoint shows you the result. The training logs and recipes show you how the team got there.

That’s what caught my eye in IFM’s xLLM release. I’d like to see what people can reproduce from the K2 Horizon materials they’ve shared.

[Orikan @GesoraMeshack]

The part of xLLM I’d test first is the data pipeline.

If changing a tokenizer or chat template really is a config change instead of another dataset reprocessing run, that could save a lot of frustrating work during model development. The repo is up, so people can dig into it.

[Lena @_LenaHart]

@GesoraMeshack I’d start with a small run and change the mixture.

[Lena @_LenaHart]

@KaiteeShiks I’d read the configuration alongside the logs.

[Kaitee @KaiteeShiks]

The K2 Horizon example made xLLM’s flexibility easier for me to understand: moving from 8K to 512K context involved changing context parallelism from 1 to 4 and micro batch size from 16 to 1.

I’d be curious how that approach holds up across other training runs.

[Antonia ziza @rafaelaziza]

produtinhos da shopee que vão salvar o seu banheiro

[Lena @_LenaHart]

@rafaelaziza só coisa linda

[Antonia ziza @rafaelaziza]

pov: você não é rica, só aprendeu a comprar na shopee

[Lena @_LenaHart]

longo mil vezes

[aliz ꕤ @nastysg_]

sempre fico com uma dúvida cruel se prefiro a selena com o cabelo longo ou curto

[Lena @_LenaHart]

@willniverso eu vi mas eu quis comentar mesmo assim

[Lena @_LenaHart]

era um viado feio

[Evandro Essência @mastheoswebdivo]

Pqp esses 🦌 de hj em dia perderam o senso do ridículo né mds como nn tem 1 pingo de respeito

[Lena @_LenaHart]

@hnrqsvd eu vi

[destiny @destinycxoh]

Selena Gomez
Kiss & Tell era aesthetic

[Lena @_LenaHart]

anotem que esse sorteio eu já ganhei porra, é meu 💉💉✨💚

[Lena @_LenaHart]

eu teria deixado só a selena e a garota

[sciuretta @sciuretta]

Scusa bambina e scusa Benny ma Selena è una figa atomica e ho dovuto cancellarvi

[Pedro Ronchi @PedroRonchi2]

O Brasil quer saber.

[Ere Pereira @ere_pereira]

A GLOBO NÃO VAI MOSTRAR ISSO VAMOS REPOSTAR

[Lena @_LenaHart]

a melhor personagem, deviam ter feito um spin off dela

[El gato Perez @El_gato_perez_]

Sabiam dessa?