Handle multiple dimensions in metal QMM + two fixes. (#2097)

This commit is contained in:
Laurent Mazare
2024-04-20 18:55:45 +02:00
committed by GitHub
parent 9215e9ce8c
commit dd78422701
2 changed files with 28 additions and 22 deletions

View File

@ -152,9 +152,9 @@ impl QMetalStorage {
// We always use a single batch dimension and stack all the tensors in the batch on the
// second dimension as the implementation in candle-metal-kernels doesn't handle batch
// properly.
let (b, m) = match dst_shape.len() {
3 => (1, dst_shape[0] * dst_shape[1]),
2 => (1, dst_shape[0]),
let m = match dst_shape.len() {
3 => dst_shape[0] * dst_shape[1],
2 => dst_shape[0],
n => crate::bail!("Invalid rank {n} for quantized matmul metal"),
};
let last_k = dst_shape.pop().unwrap();
@ -166,18 +166,23 @@ impl QMetalStorage {
let device = storage.device().clone();
let dst = device.new_buffer(dst_shape.elem_count(), DType::F32, "qmatmul")?;
let command_buffer = device.command_buffer()?;
candle_metal_kernels::call_quantized_matmul_t(
device.device(),
&command_buffer,
device.kernels(),
self.dtype.into(),
(b, m, n, k),
storage.buffer(),
layout.start_offset() * storage.dtype().size_in_bytes(),
&self.buffer,
&dst,
)
.map_err(MetalError::from)?;
// In some cases it would be better to use the mm variant, though it has its drawbacks
// around memory alignemnt.
for batch_id in 0..m {
candle_metal_kernels::call_quantized_matmul_mv_t(
device.device(),
&command_buffer,
device.kernels(),
self.dtype.into(),
(1, 1, n, k),
storage.buffer(),
(layout.start_offset() + batch_id * k) * storage.dtype().size_in_bytes(),
&self.buffer,
batch_id * n * DType::F32.size_in_bytes(),
&dst,
)
.map_err(MetalError::from)?;
}
let dst_storage = crate::MetalStorage::new(dst, device, dst_shape.elem_count(), DType::F32);
Ok((dst_storage, dst_shape))
}