Show HN: Tokiwi – An online tokenizer for any Hugging Face modeltokiwi.dev2 ポイント·投稿者 TweedBeetle·昨年·0 コメント
TweedBeetle·2 年前·議論Regarding how they’re getting to 10M context, I think it’s possible they are using the new SAMBA architecture.Here’s the paper: https://arxiv.org/abs/2312.00752And here’s a great podcast episode on it: https://www.cognitiverevolution.ai/emergency-pod-mamba-memor...
Here’s the paper: https://arxiv.org/abs/2312.00752
And here’s a great podcast episode on it: https://www.cognitiverevolution.ai/emergency-pod-mamba-memor...