How has DeepSeek improved the Transformer architecture?

Re:How has DeepSeek improved the Transformer architecture? by joshuakogut on Tuesday January 28, 2025 @04:26PM (#42858255)

Yesterday when I started evaluating Deepseek-R1 V3 it was insanely better at code generation using elaborate prompts, I asked it to write me some boilerplate code in python using the ebaysdk library to pull a list of all products sold by user with $name and it spit it out, just a few tweaks and it was ready to go.

I tried the same thing on the 7B and 32B model today, neither are as effective as codellama.

[ Parent | on HN ]

Re:How has DeepSeek improved the Transformer architecture? by ilaksh on Tuesday January 28, 2025 @07:06PM (#42859806)

I think people didn't understand my comment. I am very aware of this already.

[ Parent | on HN ]

Re:How has DeepSeek improved the Transformer architecture? by seba_dos1 on Wednesday January 29, 2025 @04:42AM (#42863201)

I think you failed to convey what you meant to with your comment.

If you want your contribution to the discussion to be meaningful, you may want to give it another go.

[ Parent | on HN ]

Re:How has DeepSeek improved the Transformer architecture? by A4ET8a8uTh0_v2 on Tuesday January 28, 2025 @06:52PM (#42859710)

I am intrigued. What did you use to run your deepseek instance?

[ Parent | on HN ]