candle
candle copied to clipboard
huggingface
Minimalist ML framework for Rust
I tried to modify the code in https://github.com/huggingface/candle/blob/main/candle-examples/examples/llama/main.rs to become a chatbot where each new prompt considers the history of all previous prompts. This is my code: ```rs use anyhow::{Error,...
Hi! I checked examples and a code. And I see only BERT implementation. Will it work with distillbert model? If not, is it hard to implement distillbert under bert?
The auto_pad attribute in candle-onnx currently only supports the default NOTSET option. It needs to be updated to include SAME_UPPER, at least, in order to run the [Tiny YOLOv3](https://github.com/onnx/models/tree/main/validated/vision/object_detection_segmentation/tiny-yolov3) model...
When I directly converted the PT file I trained into safetensors, there were no net, head, or fpn as shown in the example ``` model.0.bn.bias model.0.bn.num_batches_tracked model.0.bn.running_mean model.0.bn.running_var model.0.bn.weight model.0.conv.weight...
Perhaps we can use clamping, as per: > https://github.com/huggingface/transformers/blob/main/src/transformers/models/t5/modeling_t5.py#L748-L755 Using BF16 works on CUDA.
In Nix, builds are isolated and have limited network access. The built artifacts are written into a read-only location known as the "store". It's common to build Rust applications using...
The HuggingFace repository can be found in: https://huggingface.co/fishaudio/fish-speech-1.4
Currently, our Error type is already useful but could benefit from some of the QoL features which the anyhow crate has. This PR adds the following: - Make `Error::msg` more...
Currently `GatherElements` operator is not supported in candle-onnx. This would be convenient since a number of converted models using [HummingBird ML](https://github.com/microsoft/hummingbird) use this operator (including Tree Classifiers). `GatherElements` slightly differs...
