From 6e59aa8d81ee1a67d38293e00306230537259bff Mon Sep 17 00:00:00 2001 From: Meng Zhang Date: Thu, 26 Oct 2023 06:02:30 -0700 Subject: [PATCH] docs: fix title for the decoding blog --- website/blog/2023-10-21-incremental-decoding/index.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/website/blog/2023-10-21-incremental-decoding/index.md b/website/blog/2023-10-21-incremental-decoding/index.md index bcfc5efba49..088e57b755e 100644 --- a/website/blog/2023-10-21-incremental-decoding/index.md +++ b/website/blog/2023-10-21-incremental-decoding/index.md @@ -5,7 +5,7 @@ tags: [tech design] image: ./twitter-decoding.png --- -# Decoding the Decoding in Tabby +# Decode the Decoding in Tabby In the context of the Transformer model, which is widely used across LLMs, ***decoding*** refers to the process of generating an output sequence from an encoded input. Tabby recenty [implemented ***incremental decoding***](https://github.com/TabbyML/tabby/pull/491) as part of the greedy search. This blog will explain our thoughts behind this 🛠️💡. @@ -69,4 +69,4 @@ Incremental decoding: ......, 207, 211 -> "......[ hello]" ✅ For interested folks, you can refer to Tabby's exact implementation in `IncrementalDecoding` funcion in [`creates/tabby-inference/src/decoding.rs`](https://github.com/TabbyML/tabby/pull/491). -Have you found our new decoding methods effective? Share your thoughts with us in our [Slack](https://join.slack.com/t/tabbyml/shared_invite/zt-22thejc0z-7ePKeWNCHPX31pEtnT4oYQ) channel 🌍😊! \ No newline at end of file +Have you found our new decoding methods effective? Share your thoughts with us in our [Slack](https://join.slack.com/t/tabbyml/shared_invite/zt-22thejc0z-7ePKeWNCHPX31pEtnT4oYQ) channel 🌍😊!