de4dee

• Uploaded my Qwen3.6 27B based fine tune, after two years of experience fine tuning models

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 14 comments

• Efficient pretraining with token superposition by Nous Research

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 15 comments

• The World Inside Neural Networks

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 0 comments

• A new fine tune of Gemma 3 27B with more beneficial knowledge

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 14 comments

• Qwen 3 is better than prev versions

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 44 comments

• The AHA Indicator

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 6 comments

• Accurate representation of AI industry

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 5 comments

• No startups allowed!

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 4 comments

• Truthful AI

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 13 comments

• What is the best way and tools for converting base models to instruct?

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 5 comments

• Best roleplay model based on llama 3.1 8B?

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 2 comments

• A curated model based on my beliefs

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 18 comments

• How might LLMs store facts

  Posted by de4dee@reddit | LocalLLaMA | View on Reddit | 15 comments