Eric Buehler

Hello all and thanks for your great work here! I am integrating support for [pyo3_special_method_derive](https://crates.io/crates/pyo3_special_method_derive) into mistral.rs, but there is a linking error. The linking error seems to only occur...

Implement the Starcoder 2 model architecture

1

- [x] Loader and model - [ ] ISQ - [ ] AnyMoE - [ ] Device Mapping - [ ] X-LoRA/LoRA - [ ] Adapter activation

This is a tracking issue for the development of AnyMoE, which will be broken up into several PRs. - [x] Core functionality, plain models, all APIs: #476 - [x] Support...

Implement GPTQ quantization

1

This PR adds GPTQ quantization ([paper here](https://arxiv.org/abs/2210.17323)) support. Refs: #418, #448.

ISQ in parallel for Metal, too

1

Because #492 ensures that ISQ now happens on the CPU, we can try to run ISQ in parallel for Metal, too. Tagging: @Luddo183, this refs #393.

Use pyo3_special_method_derive to improve magic methods in Python API

1

https://crates.io/crates/pyo3_special_method_derive

Fix clippy shadowing

3

Should reduce memory usage and hopefully increase speed...

Add Rust convenience macros for loading

1

```rust use mistralrs::{device, load_normal_model}; fn main() -> anyhow::Result { let dev = Device::cuda_if_available(0)?; let runner = load_normal_model!( id = "mistralai/Mistral-7B-Instruct-v0.1".to_string(), kind = Mistral, device = dev, use_flash_attn = false );...