candle
candle copied to clipboard
huggingface
Minimalist ML framework for Rust
SYSTEM: jetson LINUX (Ubuntu 20.04.6 LTS) GitHub: https://github.com/rustai-solutions/candle_demo_yolov8 For some reason,[ jetson platform does not support nvidia-smi ](https://forums.developer.nvidia.com/t/nvidia-smi-cant-be-founded/200658/3) Maybe panic when nvidia-smi cannot be found is a bad idea? output:...
there is multiple data type tensors in the quantized models(fp16, int32..), but `candle_nn::VarBuilder` only use same dtype to load all tensors. test with [llama awq](https://huggingface.co/TheBloke/Llama-2-13B-chat-AWQ) eg: ```rust #[test] fn test_varbb()...
Flip operation similar to pytorch https://pytorch.org/docs/stable/generated/torch.flip.html
Hello. This was reported wrongly here https://github.com/EricLBuehler/candle-vllm/issues/25 but it's actually an issue with candle-core. Here is a Dockerfile reproducing the problem: ```Dockerfile # syntax=docker/dockerfile:1 # Note: if building on a...
Candle provides an example of YOLOv8, which is very useful to use. But I don't know how to train on my own dataset? Can handle directly load the model trained...
cargo run --example resnet --release --features metal -- --image test.jpg help to solve this problem! Thanks!
### Changes - The input stream from the microphone is built once in the thread that handles it. - Inside the loop block, the stream is played and paused for...
The current implementation of key-value concatenation relies on Tensor::cat which uses ucopy kernel to compute every indice for all output tensor elements based on the input shape and output stride...
I am working on some project that started as `tch-rs`, and would like to know if there is any way to convert a `tch-rs::Tensor` as a `candle_core::Tensor`? preferably zero-copy?
