From 0d4d18df51eae86dbd18140f634928715e94f9ce Mon Sep 17 00:00:00 2001 From: whitney Date: Thu, 23 Jan 2025 14:38:25 -0800 Subject: [PATCH] Update deepseek blog --- website/blog/2025-01-22-pure-rl-with-deepseek/index.mdx | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/website/blog/2025-01-22-pure-rl-with-deepseek/index.mdx b/website/blog/2025-01-22-pure-rl-with-deepseek/index.mdx index 9a8a077..60c97f0 100644 --- a/website/blog/2025-01-22-pure-rl-with-deepseek/index.mdx +++ b/website/blog/2025-01-22-pure-rl-with-deepseek/index.mdx @@ -16,6 +16,12 @@ community has really been putting in the work so those like me looking to start [OpenWebUI](https://github.com/ml-explore/OpenWebUI) is a great frontend UI for interacting with models. Some of their docker containers come bundled with Ollama, which means the setup is literally as simple as building a super standard docker container with it. +OpenWebUI also has the ability to allow for Web Search. In a departure from previous attempts, it seems to be possible to throw together +a few free and/or open source components together to make a compelling attempt at web search. + +Because DeepSeek responds in a very specific and waffle-y way, it's not very well suited for being the model we use to generate our +search query. In its place, Llama3.2 is a suitable replacement as it's better with direct instructions like our ask to make the query. + # Giving it a Shot ## DeepSeek