Source code for modelzoo.transformers.pytorch.layers_api_demo.cb_transformer

# Copyright 2022 Cerebras Systems.
#
# 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.

import math

from torch import Tensor, nn

from modelzoo.common.pytorch.layers import (
    EmbeddingLayer,
    TransformerEncoder,
    TransformerEncoderLayer,
)


class TransformerModel(nn.Module):
    def __init__(
        self,
        ntoken: int,
        d_model: int,
        nhead: int,
        d_hid: int,
        nlayers: int,
        dropout: float = 0.5,
        activation: str = "gelu",
        max_len: int = 10,
    ):
        super().__init__()
        self.model_type = "Transformer"

        self.encoder = EmbeddingLayer(
            vocab_size=ntoken,
            embedding_size=d_model,
            position_embedding_type="fixed",
            max_position_embeddings=max_len,
        )

        encoder_layers = TransformerEncoderLayer(
            d_model,
            nhead,
            d_hid,
            dropout,
            batch_first=True,
            activation=activation,
        )
        self.transformer_encoder = TransformerEncoder(encoder_layers, nlayers)

        self.d_model = d_model
        self.decoder = nn.Linear(d_model, ntoken)

        self.init_weights()

    def init_weights(self) -> None:
        initrange = 0.1
        self.decoder.bias.data.zero_()
        self.decoder.weight.data.uniform_(-initrange, initrange)

    def forward(self, src: Tensor, src_mask: Tensor) -> Tensor:
        """
        Args:
            src: Tensor, shape [seq_len, batch_size]
            src_mask: Tensor, shape [seq_len, seq_len]
        Returns:
            output Tensor of shape [seq_len, batch_size, ntoken]
        """
        src = self.encoder(src) * math.sqrt(self.d_model)
        src = self.transformer_encoder(src, src_mask)
        output = self.decoder(src)
        return output