ngxson/wllama

Issues

• What does `noTEE` do?

  #107 opened a month ago by flatsiedatsie
  0

• error loading model hyperparameters

  #106 opened a month ago by flatsiedatsie
  3

• Unreachable

  #62 opened 3 months ago by flatsiedatsie
  4

• [Feature request] LoRA support

  #105 opened a month ago by OKUA1
  0

• implement KV cache reuse for completion

  #101 opened 2 months ago by ngxson
  0

• main: initialize main example

  #96 opened 2 months ago by ngxson
  2

• Add prettier

  #98 opened 2 months ago by ngxson
  0

• Add support for `AbortController` on downloading model

  #83 opened 3 months ago by flatsiedatsie
  1

• ci: add e2e test

  #97 opened 2 months ago by ngxson
  0

• T5 and Flan-T5 models support (llama_encode)

  #86 opened 2 months ago by felladrin
  1

• Model caching with new download manager?

  #87 opened 2 months ago by flatsiedatsie
  1

• Add support for control vectors

  #89 opened 2 months ago by ngxson
  0

• BitNet support

  #69 opened 2 months ago by flatsiedatsie
  17

• Should all models now be chunked?

  #20 opened 2 months ago by flatsiedatsie
  3

• The mystery of Schrodinger's exit function

  #82 opened 3 months ago by flatsiedatsie
  2

• Failed to build from scratch: llamacpp-wasm-builder, CMake Error (add_executable): Cannot find source file

  #76 opened 3 months ago by flatsiedatsie
  1

• Large models fail to load from cache on iOS browsers, but load and run fine when uncached

  #72 opened 3 months ago by felladrin
  5

• Feature request: Github build workflow

  #6 opened 5 months ago by flatsiedatsie
  1

• Add WebGPU support

  #66 opened 3 months ago by ngxson
  0

• PostMessage: Data cannot be cloned, out of memory

  #12 opened 4 months ago by flatsiedatsie
  23

• After upgrading to version 1.8.0, the async function `loadModelFromUrl` is not completing when using large models

  #31 opened 4 months ago by felladrin
  6

• Feature request: diversify error mesages when loading a model fails

  #56 opened 4 months ago by flatsiedatsie
  1

• unlimited token limit in demo

  #71 opened 3 months ago by fabriziosalmi
  2

• Glitch remixable no-build example

  #70 opened 3 months ago by Utopiah
  2

• [Idea] Use OPFS for storing downloaded files

  #38 opened 3 months ago by ngxson
  1

• Would it be possible to release a new version?

  #67 opened 3 months ago by flatsiedatsie
  5

• Error when loading a model via relative path

  #63 opened 3 months ago by felladrin
  3

• Made a function to build the Model URL Array when detecting the url has the gguf-split pattern `-<number>-of-<number>.gguf`. Would it fit in the lib?

  #58 opened 3 months ago by felladrin
  2

• 404-ed model still ended up in wllama_cache?

  #57 opened 4 months ago by flatsiedatsie
  3

• [Idea] Load model from File Blob

  #42 opened 4 months ago by ngxson
  0

• [Idea] Stream data from main thread to worker

  #43 opened 4 months ago by ngxson
  1

• How would you implement RAG / Document chat?

  #36 opened 4 months ago by flatsiedatsie
  5

• Post on Reddit/r/LocalLlama?

  #53 opened 4 months ago by flatsiedatsie
  8

• [Idea] Publish to JSR

  #55 opened 4 months ago by ngxson
  0

• Error when running `h2o-danube2-1.8b-chat` and `phi-2` models when `cache_type_k` is set to `q4_0` or `q8_0`

  #54 opened 4 months ago by felladrin
  3

• warning: munmap failed: Invalid argument

  #7 opened 4 months ago by flatsiedatsie
  6

• Feature request: Add an option for Verbose Mode, for opt-in console logs

  #17 opened 4 months ago by felladrin
  1

• Feature request: get the debug output through a callback?

  #40 opened 4 months ago by flatsiedatsie
  1

• Seeing <|end|> in output

  #45 opened 4 months ago by flatsiedatsie
  4

• performance expectations

  #4 opened 5 months ago by chadkirby
  5

• missing pre-tokenizer type

  #41 opened 4 months ago by flatsiedatsie
  11

• [Idea] Use something better than memfs

  #35 opened 4 months ago by ngxson
  2

• Wllama doesn't load the provided chunks

  #44 opened 4 months ago by flatsiedatsie
  2

• Bug: exception handling is broken

  #22 opened 4 months ago by ngxson
  0

• Feature request: add a built-in way to interrupt inference

  #19 opened 4 months ago by flatsiedatsie
  0

• Out Of Memory error in Wllama with multi-threads on iOS browser

  #18 opened 4 months ago by felladrin
  4

• The current configuration of Emscripten with `PTHREAD_POOL_SIZE=32` for multi-threading may be causing memory wastage

  #16 opened 4 months ago by felladrin
  3

• qwen returns empty string

  #11 opened 4 months ago by flatsiedatsie
  4

• null function or function signature mismatch at wllama.wasm

  #10 opened 4 months ago by flatsiedatsie
  5

• Support for local webpage use?

  #5 opened 5 months ago by twoxfh
  2