utils.py

# coding=utf-8
# Copyright (c) 2019, NVIDIA CORPORATION.  All rights reserved.
#
# 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.

"""Utilities for models."""

import math

import torch

from .transformer import LayerNorm


def init_method_normal(sigma):
    """Init method based on N(0, sigma)."""
    def init_(tensor):
        return torch.nn.init.normal_(tensor, mean=0.0, std=sigma)

    return init_


def scaled_init_method_normal(sigma, num_layers):
    """Init method based on N(0, sigma/sqrt(2*num_layers)."""
    std = sigma / math.sqrt(2.0 * num_layers)

    def init_(tensor):
        return torch.nn.init.normal_(tensor, mean=0.0, std=std)

    return init_


def get_linear_layer(rows, columns, init_method):
    """Simple linear layer with weight initialization."""
    layer = torch.nn.Linear(rows, columns)
    init_method(layer.weight)
    with torch.no_grad():
        layer.bias.zero_()
    return layer


@torch.jit.script
def gelu_impl(x):
    """OpenAI's gelu implementation."""
    return 0.5 * x * (1.0 + torch.tanh(0.7978845608028654 * x *
                                       (1.0 + 0.044715 * x * x)))


def gelu(x):
    return gelu_impl(x)


def get_params_for_weight_decay_optimization(module):
    """Divide params into with-weight-decay and without-weight-decay groups.
    Layernorms and baises will have no weight decay but the rest will.
    """
    weight_decay_params = {'params': []}
    no_weight_decay_params = {'params': [], 'weight_decay': 0.0}
    for module_ in module.modules():
        if isinstance(module_, LayerNorm):
            no_weight_decay_params['params'].extend(
                [p for p in list(module_._parameters.values())
                 if p is not None])
        else:
            weight_decay_params['params'].extend(
                [p for n, p in list(module_._parameters.items())
                 if p is not None and n != 'bias'])
            no_weight_decay_params['params'].extend(
                [p for n, p in list(module_._parameters.items())
                 if p is not None and n == 'bias'])

    return weight_decay_params, no_weight_decay_params