Add Quantized Ops to the JIT (#1204)

* JIT for quantized ops

* remove unused imports

* address comments

* fix imports

* second attempt to fix imports

---------

Co-authored-by: Alex Barron <abarron22@apple.com>
This commit is contained in:
Alex Barron
2024-06-12 09:47:12 -07:00
committed by GitHub
parent df964132fb
commit dd7d8e5e29
13 changed files with 1778 additions and 1948 deletions

View File

@@ -195,13 +195,16 @@ MTL::ComputePipelineState* get_fft_kernel(
metal::Device& d,
const std::string& kernel_name,
const std::string& hash_name,
const int tg_mem_size,
const std::string& in_type,
const std::string& out_type,
int step,
bool real,
const metal::MTLFCList& func_consts) {
const metal::MTLFCList& func_consts,
const std::string&) {
return d.get_kernel(kernel_name, "mlx", hash_name, func_consts);
}
MTL::ComputePipelineState* get_quantized_kernel(
metal::Device& d,
const std::string& kernel_name,
const std::string&) {
return d.get_kernel(kernel_name);
}
} // namespace mlx::core