Skip to content

Commit

Permalink
Apply quantization on megablox kernel; support both training and serving
Browse files Browse the repository at this point in the history
  • Loading branch information
lenscloth committed Dec 16, 2024
1 parent 1e39608 commit 7878fca
Show file tree
Hide file tree
Showing 6 changed files with 1,072 additions and 23 deletions.
16 changes: 16 additions & 0 deletions MaxText/kernels/megablox/__init__.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
# Copyright 2024 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""Megablox kernel"""

from kernels.megablox.ops import gmm
59 changes: 59 additions & 0 deletions MaxText/kernels/megablox/common.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,59 @@
# Copyright 2024 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""Common utilities for GMM kernels."""

import re

import jax
import jax.numpy as jnp


def is_tpu() -> bool:
return "TPU" in jax.devices()[0].device_kind


def tpu_kind() -> str:
"""Query identification string for the currently attached TPU."""
return jax.devices()[0].device_kind


_TPU_KIND_PATTERN = re.compile(r"TPU v(\d+)")


def tpu_generation() -> int:
"""Generation number of the currently attached TPU."""
if version := _TPU_KIND_PATTERN.match(tpu_kind()):
return int(version[1])
raise NotImplementedError("only TPU devices are supported")


def supports_bfloat16_matmul() -> bool:
"""Does the currently attached CPU support bfloat16 inputs?"""
return not is_tpu() or tpu_generation() >= 4


def assert_is_supported_dtype(dtype: jnp.dtype) -> None:
if dtype not in (jnp.bfloat16, jnp.float32):
raise ValueError(f"Expected bfloat16 or float32 array but got {dtype}.")


def select_input_dtype(lhs: jnp.ndarray, rhs: jnp.ndarray) -> jnp.dtype:
"""A type to which both input should be adapted to before dot product."""
# bf16xbf16 matmul is only supported since TPUv4 generation. In case of mixed
# input precision, we need to convert bf16 argument to fp32 beforehand.
if supports_bfloat16_matmul() and lhs.dtype == jnp.bfloat16 and rhs.dtype == jnp.bfloat16:
return jnp.bfloat16
else:
return jnp.float32
Loading

0 comments on commit 7878fca

Please sign in to comment.