LiquidGunay

• Best way to add "Memory" to LLMs?

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 62 comments

• Chatkit-js with LangGraph Agents?

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

• Fastest way to serve llama 3 8b

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 33 comments

• Is the TPU really an ASIC?

  Posted by LiquidGunay@reddit | hardware | View on Reddit | 56 comments

• No AWQ for Gemma 3?

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 26 comments

• GRPO for VLMs?

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

• The real use case for DIGITS is SLM training

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 16 comments

• How to make Coding LMs more creative?

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 8 comments

• Is Mamba inference faster than Transformers? (in practice)

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 8 comments

• What happened to the Nvidia VLM?

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

• Is serving a quantized model faster?

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

• Hard RAG benchmarks?

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

• Task specific fine-tuning using distillation?

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

• What architectural changes would be required to make an omni model?

  Posted by LiquidGunay@reddit | LocalLLaMA | View on Reddit | 11 comments

• If companies were waifus

  Posted by LiquidGunay@reddit | Jokes | View on Reddit | 2 comments