Enn

Hi, im Tuan

[Enn @teeyouaenn]

@itswillhua Also miss replying to human reviewers 😭

[Enn @teeyouaenn]

@demisama_ 2 failed road tests 😭😭

[Enn @teeyouaenn]

@_Wienman_ @asecretcrow Second!

[Enn @teeyouaenn]

@KevinKaichuang @eLife @SarahAlamdari @nityathakkar_ @vdbergrianne @alexijielu @AMosesLab @nfusi @avapamini Congrats!

[Enn @teeyouaenn]

@chang_m_yun Oh that makes more sense; but ive seen some labs operate on 20x 4090s hahahaha

[Enn @teeyouaenn]

@chang_m_yun Has “a few h100s and h200s” and you say its not enough…

[Enn @teeyouaenn]

@chang_m_yun Pre-Training on test set is all you need…

[Enn @teeyouaenn]

@chang_m_yun @GenomeBrowser Congrats!

[Enn @teeyouaenn]

@ZhengyangG51844 I wonder if this is one way to achieve it; sampling from latent space. Or can design a very specific probalistic or energy-based method to distribute over an arbitrary amount of output length

[Enn @teeyouaenn]

@chang_m_yun Hey, im back! Sorry, still have no idea how X works… I wonder if we can use evo2 sequence embedding and co-train it with epigenomic signals and see what will happen (similar to Evo2HiC). Variant prediction across distal sequences is an interesting challenge

[Enn @teeyouaenn]

@chang_m_yun Love to see efforts benchmarking! One perspective I always have is that s2f models are always trained reference genomes, and it doesn’t really make sense to do variant prediction. Maybe there should be a concerted effort for perturbseq/crispr assays to improve this capability

[Enn @teeyouaenn]

@ZhengyangG51844 Hi, ive sent you a dm! Lmk if it went through

[Enn @teeyouaenn]

@sheriyuo Can i join Stepfun; sounds like fun 😄

[Enn @teeyouaenn]

@pkoo562 Would love to help!

[Enn @teeyouaenn]

@vaishsingh_ That sounds cool! May I attend

[Enn @teeyouaenn]

@sheriyuo This is good qs well https://t.co/AQDupqaby0

[Enn @teeyouaenn]

@sheriyuo Go for it!

[Enn @teeyouaenn]

@kimnoel Followed

[Enn @teeyouaenn]

@ShitpostGate This is not remotely funny