≪ View Page

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
infrastructure:llama-cpp-on-dreamflasher [2026/10/10 22:19] – diamondinfrastructure:llama-cpp-on-dreamflasher [2026/10/10 22:32] (current) – diamond
Line 2: Line 2:
  
 - admin: @diamond - admin: @diamond
-- address: https://llama-cpp.dreamflasher.dma.space/+- addresses: 
 +  - external: https://llama-cpp.dreamflasher.dma.space/ (login required) 
 +  - internal: http://dreamflasher.int.dma.computer:12345/
 - deployment: https://codeberg.org/dma/infra/src/commit/main/nixos/machines/dreamflasher/_llama-server.nix - deployment: https://codeberg.org/dma/infra/src/commit/main/nixos/machines/dreamflasher/_llama-server.nix
 +
 +## Inference Engine
 +
 +- tool: llama.cpp
 +- model name: Qwen/Qwen3.6-35B-A3B (`Qwen3.6-35B-A3B-UD-IQ2_XXS.gguf`)
 +- model size: 10.01 GB
 +- context size: 131,072 tokens
 +- training context: 262,144 tokens
 +- parameters: 34.66B
 +- embedding size: 2,048
 +- vocabulary size: 248,320 tokens
 +- parallel slots: 4
  
 ## Usage Guidelines ## Usage Guidelines
Line 14: Line 28:
  
 > **Note**: We don't store chat history. Inference is done completely offline. > **Note**: We don't store chat history. Inference is done completely offline.
 +
 +