Stas Bekman

Results 664 comments of Stas Bekman

Thank you for sharing the working code, @aloy99! I was able to reproduce it. So you're saying it's lower than the paper, I run just the `flare_finqa` task and got...

Shouldn't you compare it against their latest paper which in theory should match the source code on this repo? https://arxiv.org/abs/2402.12659 or perhaps the original paper you linked to should match...

Finally got the time to read the paper, added it here: https://github.com/stas00/ml-engineering/commit/185f7a59918e1782b43eca45b10cd0a37ed56559 Thank you for the recommendation, @boweiliu - a fantastic paper!

@yaolu, indeed that was a great technical post. Let's add it as well https://github.com/stas00/ml-engineering/commit/a7ef3a63f7a8ff32b88968509c0202244a2bad65

I'm not sure how `--enable_cpu_affinity` can help with this issue and when it does the improvement is very small. But I see that there an indepth discussion of this particular...

Hmm, but if the intention is to be pip-BC `-U` == update only listed packages - is there a reason to break that well-established pattern?

I suppose most of the time we don't care if `-U` updates all packages, except when there is a situation where some package is either broken or leads to a...

> I also need to look at what happens with `pip install -U -r requirements.txt` -- does it upgrade all the listed packages? it appears to be so: ``` $...

Indeed, recursive updates related to the listed packages - not all installed packages - we presume that is since the writing is ambiguous, but we can derive their truth/reality from...

I did: ``` HuggingFaceDatasetReader( dataset="HuggingFaceFW/fineweb", dataset_options=dict(name="CC-MAIN-2024-10", split="train"), streaming=True, # to avoid the huge overhead of memory and time limit=10000, batch_size=1, doc_progress=True, text_key="text", ), ``` I suspect the prompt comes from...