Stas Bekman

Results 664 comments of Stas Bekman

Done: https://github.com/huggingface/datasets/issues/6614

Are you asking about https://github.com/stas00/porting/tree/master/datasets/pegasus I think a lot of this data has been since ported to https://huggingface.co/datasets e.g.: https://huggingface.co/datasets/billsum so it should be easy to convert to .csv using...

thanks, @sagar-deepscribe - that was helpful. Here is your code slightly edited and w/ copy-n-paste instructions to run: ``` cat vllm-lora-convert.py import sys import safetensors.torch src, dst = sys.argv[-2:] tensors...

Very odd. I have even rechecked with incognito mode, with all extensions turned off, it's not there. And since chrome renders it fine on the same account, then this is...

By trying incognito I meant to signal that it's not due to an impact of some FF extension, since incognito disables them all. But if it's just me, and you...

Also should probably assert if `int4` attempted to be used w/o `deepspeed>=xyz` once the DS PR is merged... could tentatively set to the next deepspeed version? perhaps with XXX to...

OK, I think I understand the limitations of pytorch and it'll get only worse when you try `int3`, etc. even if `int4` is supported. https://github.com/huggingface/transformers-bloom-inference/pull/37/files#r1026981222 I propose we break the...

may I suggest that the just added `kwargs['quant']['weight']['num_bits']` isn't the most user-friendly API as far as `kwargs` go? why not have a flat structure of simple key=value pairs and once...

>Even though this approach is extremely intrusive, it does have its merits of support a wider range of models + being a bit more expressive. I would relate this to...