File size: 2,353 Bytes
9c6594c |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
# Copyright The Lightning AI team.
#
# 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
#
# http://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.
from contextlib import AbstractContextManager
from typing import Any, Literal
import torch
from lightning_utilities.core.apply_func import apply_to_collection
from torch import Tensor
from torch.nn import Module
from typing_extensions import override
from lightning_fabric.plugins.precision.precision import Precision
from lightning_fabric.plugins.precision.utils import _convert_fp_tensor, _DtypeContextManager
class HalfPrecision(Precision):
"""Plugin for training with half precision.
Args:
precision: Whether to use ``torch.float16`` (``'16-true'``) or ``torch.bfloat16`` (``'bf16-true'``).
"""
precision: Literal["bf16-true", "16-true"] = "16-true"
def __init__(self, precision: Literal["bf16-true", "16-true"] = "16-true") -> None:
self.precision = precision
self._desired_input_dtype = torch.bfloat16 if precision == "bf16-true" else torch.float16
@override
def convert_module(self, module: Module) -> Module:
return module.to(dtype=self._desired_input_dtype)
@override
def tensor_init_context(self) -> AbstractContextManager:
return _DtypeContextManager(self._desired_input_dtype)
@override
def module_init_context(self) -> AbstractContextManager:
return self.tensor_init_context()
@override
def forward_context(self) -> AbstractContextManager:
return self.tensor_init_context()
@override
def convert_input(self, data: Any) -> Any:
return apply_to_collection(data, function=_convert_fp_tensor, dtype=Tensor, dst_type=self._desired_input_dtype)
@override
def convert_output(self, data: Any) -> Any:
return apply_to_collection(data, function=_convert_fp_tensor, dtype=Tensor, dst_type=torch.get_default_dtype())
|