Pinned to Ideas

• 13 You must be logged in to vote

  💡

  GPU optimization across different cards

  performance Speed related topics hardware Hardware related

  JohannesGaessler started May 13, 2023 in Ideas

  16

Discussions

• 2 You must be logged in to vote

  💡

  Weight Layer subsystem

  mwt11a started Sep 23, 2023 in Ideas

  

  

  

  

  7
• 3 You must be logged in to vote

  💡

  Loader for parameter configuration profiles

  tomsnunes started Apr 1, 2023 in Ideas

  

  

  1
• 4 You must be logged in to vote

  💡

  Can we use both the CPU and GPU(and may the NE) on unified memory systems(Mac)?

  JRZS started Sep 8, 2023 in Ideas

  

  

  

  

  14
• 3 You must be logged in to vote

  💡

  Fine-tune LLM with llama.cpp community data to enhance search efficiency

  bobqianic started Sep 15, 2023 in Ideas

  

  

  1
• 12 You must be logged in to vote

  💡

  Support IBM MoLM. Possible nextgen llm architecture

  Jipok started Sep 14, 2023 in Ideas

  

  

  0
• 1 You must be logged in to vote

  💡

  Grammar conditional output

  eamag started Sep 11, 2023 in Ideas

  

  

  1
• 6 You must be logged in to vote

  💡

  Medusa support for faster decoding

  Kimiko-AI started Sep 12, 2023 in Ideas

  

  

  0
• 5 You must be logged in to vote

  💡

  Unlimiformer

  enhancement New feature or request research 🔬

  xaedes started May 7, 2023 in Ideas

  

  

  

  

  21
• 1 You must be logged in to vote

  💡

  (Falcon 180B) Flexgen throughput strategy w/ everyman desktop combo + Quant's??

  BarfingLemurs started Sep 6, 2023 in Ideas

  

  

  

  

  5
• 2 You must be logged in to vote

  💡

  Enhance grammar by supporting Regex's curly bracket quantifiers

  Mihaiii started Sep 4, 2023 in Ideas

  

  

  2
• 2 You must be logged in to vote

  💡

  Table of Best Results

  SilvaRaulEnrique started Aug 24, 2023 in Ideas

  

  

  

  

  6
• 1 You must be logged in to vote

  💡

  Generate a list of the most likely answers

  nickfreeman-de started Aug 25, 2023 in Ideas · Closed

  

  

  

  

  4
• 4 You must be logged in to vote

  💡

  Root mean square of token probability differences as new quantization quality metric

  JohannesGaessler started Aug 29, 2023 in Ideas

  

  

  

  

  10
• 2 You must be logged in to vote

  💡

  Will it be possible to use speculative sampling in llama.cpp?

  qnixsynapse started Aug 28, 2023 in Ideas

  

  

  2
• 3 You must be logged in to vote

  💡

  "Arthur Bench: The Most Robust Way to Evaluate LLMs"

  ianscrivener started Aug 27, 2023 in Ideas

  

  

  2
• 5 You must be logged in to vote

  💡

  Consider the use of semver tags for builds

  longregen started Aug 23, 2023 in Ideas

  

  

  0
• 2 You must be logged in to vote

  💡

  I know this would require a new model structure, or maybe not, but hear me out.

  philtimmes started Aug 14, 2023 in Ideas

  

  

  

  

  5
• 7 You must be logged in to vote

  💡

  QuIP: 2-Bit Quantization of Large Language Models With Guarantees

  research 🔬

  abceleung started Aug 4, 2023 in Ideas

  

  

  

  

  14
• 1 You must be logged in to vote

  💡

  Enable CPU HBM

  jikunshang started Aug 14, 2023 in Ideas

  

  

  1
• 1 You must be logged in to vote

  💡

  A sequence repetition penalty sampler - how would you expect it to work?

  KerfuffleV2 started Aug 10, 2023 in Ideas · Closed

  

  

  0
• 11 You must be logged in to vote

  💡

  Request: add support for Cerebras GPT models just released

  🦙. llama model Model specific generation quality Quality of model output

  Topping1 started Mar 28, 2023 in Ideas

  

  

  

  

  10
• 1 You must be logged in to vote

  💡

  Implementation of ReAct framework

  superchargez started Jul 31, 2023 in Ideas

  

  

  0
• 3 You must be logged in to vote

  💡

  Setting logit biases as you go instead of just at start.

  superkuh started Jul 30, 2023 in Ideas

  

  

  0
• 14 You must be logged in to vote

  💡

  bark.cpp? New TTS model released

  r3nor started Apr 20, 2023 in Ideas

  

  

  

  

  7
• 1 You must be logged in to vote

  💡

  Add differents way to load a model

  ostix360 started Jul 26, 2023 in Ideas

  

  

  2