Elvir Crnčević
Elvir Crnčević
**Describe the bug** After downloading the linux version and running the binary, I get the following error: ` SDL error opening audio device: Audio subsystem is not initialized ` **To...
Hi, I hope that this repo is still maintained or at least open for questions. :) My use-case: I have a code-base which utilizes the C++ wmma template API. For...
This PR adds support for the following: - Efficient SPQR CUDA-based matvec kernel implementation for a subset of paramaters - Integration of said kernel for end-to-end inference - Kernel benchmarks...
Adds static storage size check during move assignments.
The purpose of this PR is to track the following features: * Optimize 3-bit dequanzation. * Add support for multi-batch inference. * Add support for efficient spqr-to-dense matrix dequantization. *...
The PR contains important fixes for: * proper device placement * position_ids being None exception related to the updated transformers package * make seqlen a passable parameter to main.py *...