candle
candle copied to clipboard
huggingface
Minimalist ML framework for Rust
The m2 metal support is great for people running macs, however it doesn't cover the majority of devices out there. Supporting the Vulkan graphics library can greatly increase the reach...
Hi, This library is cool. Rust for deep learning is nice and great work from huggingface. I am curious to understand if there are plans for AMD hardware support for...
Out of the box, Windows does not provide two things required by the CUDA feature flag: 1. `cl.exe` being in PATH. 2. An Nvidia driver up-to-date enough to allow querying...
`argmax_keepdim` has no documentation, which would be good to fix in general, but especially to resolve the question of what occurs when there are multiple elements along the target dimension...
This is a reopened PR #1680, it will improve around 5% - 10% performance for every LLM model in candle-examples. This was also adopted by @EricLBuehler in his Mistralrs project...
When using the YOLO project in the handle wasm example, it takes a long time to make predictions on the browser side, and the CPU usage remains unchanged. How can...
``` C:\Users\igumn\candle\candle-examples\examples\mistral>cargo run --example mistral --features accelerate --release -- --prompt "Here is a sample quick sort implementation in rust " --quantized -n 400 Compiling cc v1.0.90 Compiling serde v1.0.197 Compiling...
Fixes for BERT Safetensors, where current HF model.safetensors has: - LayerNorm gamma instead of weight - LayerNorm beta instead of bias - bert variable prefix
This PR adds flash attention to whisper and can be run with the following command. ```bash cargo run --example whisper --release --features flash-attn -- --use-flash-attn ``` *note to enable candle...
