Matthew Honnibal
Matthew Honnibal
@mauryaland I hope we can have annotations starting in January. The first data to be annotated will be English and German, with other annotation projects hopefully starting fairly quickly. We'll...
@hanryhu 😨 Thanks for the example, that definitely looks wrong. I wonder what's going on there, hm. I doubt `salad` is even an unseen word!
I think the issue is: ``` gpu_allocator = "tensorflow" ``` We only support transformers on PyTorch currently, so you'll need to change this to `pytorch`.
We don't have that functionality yet unfortunately. I hope we can provide it in a future release.
In principle I'm in favour of this, and we've looked into it at various points. However, in practice it's a relatively difficult optimisation to carry and maintain. The general problem...
Oh thanks for explaining this! I didn't know about it. I've definitely been frustrated by Pickle before. I think there should be a way to do this cleverly if we...
I think you'll have an easier time of things if you try to parallelise over larger units of work, e.g. a few dozen megabytes of compressed text per process. I...
My position on this is mostly "keep it as is". I'm open to debate on this, but I'll explain my position. I agree that an `is_sentenced` flag would be good....
Thanks for writing this up, it's a very helpful explanation that's sure to save us time.
Okay we have it on good authority that bundling the DLL is the right thing: https://twitter.com/honnibal/status/1263388610021203968 . So we need to add that to our build process somehow. Possibly in...