amplify_te.py

# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: LicenseRef-Apache2
#
# 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.

# Adapted from https://huggingface.co/chandar-lab/AMPLIFY_120M/blob/main/amplify.py

import torch
from torch import nn
import transformer_engine.pytorch
from transformer_engine.pytorch.attention.rope import RotaryPositionEmbedding
from transformers import PretrainedConfig, PreTrainedModel
from transformers.modeling_outputs import BaseModelOutput, MaskedLMOutput


class AMPLIFYConfig(PretrainedConfig):
    model_type = "AMPLIFY"

    # All config parameters must have a default value.
    def __init__(
        self,
        hidden_size: int = 960,
        num_hidden_layers: int = 32,
        num_attention_heads: int = 15,
        intermediate_size: int = 3840,
        dropout_prob: float = 0,
        embedding_init_range: float = 0.02,
        decoder_init_range: float = 0.02,
        rms_norm: bool = True,
        norm_eps: float = 1e-05,
        hidden_act: str = "SwiGLU",
        layer_norm_after_embedding: bool = False,
        layer_norm_before_last_layer: bool = True,
        vocab_size: int = 27,
        ffn_bias: bool = False,
        att_bias: bool = False,
        pad_token_id: int = 0,
        max_length: int = 2048,
        **kwargs,
    ):
        super().__init__(**kwargs)

        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.intermediate_size = intermediate_size
        self.dropout_prob = dropout_prob
        self.embedding_init_range = embedding_init_range
        self.decoder_init_range = decoder_init_range
        self.rms_norm = rms_norm
        self.norm_eps = norm_eps
        self.hidden_act = hidden_act
        self.layer_norm_after_embedding = layer_norm_after_embedding
        self.layer_norm_before_last_layer = layer_norm_before_last_layer
        self.vocab_size = vocab_size
        self.ffn_bias = ffn_bias
        self.att_bias = att_bias
        self.pad_token_id = pad_token_id
        self.max_length = max_length


class AMPLIFYPreTrainedModel(PreTrainedModel):
    config_class = AMPLIFYConfig

    def _init_weights(self, module):
        if isinstance(module, nn.Linear):
            module.weight.data.uniform_(
                -self.config.decoder_init_range, self.config.decoder_init_range
            )
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.uniform_(
                -self.config.embedding_init_range, self.config.embedding_init_range
            )


class AMPLIFY(AMPLIFYPreTrainedModel):
    """The main model class.

    Args:
       config (amplify.model.amplify.AMPLIFYConfig): model configuration.
    """

    def __init__(self, config: AMPLIFYConfig, **kwargs):
        super().__init__(config)

        self.config = config

        self.encoder = nn.Embedding(
            config.vocab_size,
            config.hidden_size,
            padding_idx=config.pad_token_id,
            dtype=config.torch_dtype,
        )

        if config.layer_norm_after_embedding:
            self.layer_norm_1 = (
                transformer_engine.pytorch.RMSNorm(
                    config.hidden_size, config.norm_eps, params_dtype=config.torch_dtype
                )
                if config.rms_norm
                else transformer_engine.pytorch.LayerNorm(
                    config.hidden_size, config.norm_eps, params_dtype=config.torch_dtype
                )
            )

        if config.hidden_act.lower() == "swiglu":
            # To keep the number of parameters and the amount of computation constant, we reduce the
            # number of hidden units by a factor of 2/3 (https://arxiv.org/pdf/2002.05202.pdf) and
            # make it a multiple of 8 to avoid RuntimeError due to misaligned operand
            multiple_of = 8
            intermediate_size = int(2 * config.intermediate_size / 3)
            intermediate_size = multiple_of * (
                (intermediate_size + multiple_of - 1) // multiple_of
            )

        self.transformer_encoder = nn.ModuleList()
        for layer_num in range(config.num_hidden_layers):
            self.transformer_encoder.append(
                transformer_engine.pytorch.TransformerLayer(
                    hidden_size=config.hidden_size,
                    ffn_hidden_size=intermediate_size,
                    num_attention_heads=config.num_attention_heads,
                    layernorm_epsilon=config.norm_eps,
                    hidden_dropout=config.dropout_prob,
                    attention_dropout=config.dropout_prob,
                    apply_residual_connection_post_layernorm=False,
                    layer_type="encoder",
                    self_attn_mask_type="padding",
                    normalization="RMSNorm" if config.rms_norm else "LayerNorm",
                    fuse_qkv_params=True,
                    qkv_weight_interleaved=True,
                    output_layernorm=False,
                    bias=False,
                    activation=config.hidden_act.lower(),
                    attn_input_format="bshd",
                    layer_number=layer_num + 1,
                    name="encoder_block",
                    window_size=(-1, -1),
                    rotary_pos_interleaved=True,
                    seq_length=config.max_length,
                    params_dtype=config.torch_dtype,
                )
            )

        self.freqs_cis = RotaryPositionEmbedding(
            config.hidden_size // config.num_attention_heads, interleaved=True
        )(config.max_length)

        # Initialize weights and apply final processing
        self.post_init()

    def forward(
        self,
        input_ids,
        attention_mask=None,
        output_hidden_states=False,
        output_attentions=False,
        labels=None,
        **kwargs,
    ):
        # Initialize
        hidden_states = []

        # Attention mask
        if attention_mask is not None and attention_mask.dtype is torch.int64:
            # TE expects a boolean attention mask, where "True" indicates a token to be masked.
            attention_mask = ~attention_mask.to(bool)

        # RoPE
        self.freqs_cis = self.freqs_cis.to(input_ids.device, non_blocking=True)
        freqs_cis = self.freqs_cis[: input_ids.shape[1]]

        # Embedding
        x = self.encoder(input_ids)
        if self.config.layer_norm_after_embedding:
            x = self.layer_norm_1(x)

        # Transformer encoder
        for layer in self.transformer_encoder:
            x = layer(x, attention_mask, rotary_pos_emb=freqs_cis)
            if output_hidden_states:
                hidden_states.append(x)
            if output_attentions:
                raise ValueError("output_attentions is not supported for TE")

        return BaseModelOutput(
            last_hidden_state=x,
            hidden_states=hidden_states,
            attentions=None,
        )


class AMPLIFYForMaskedLM(AMPLIFYPreTrainedModel):
    def __init__(self, config: AMPLIFYConfig, **kwargs):
        super().__init__(config)
        self.amplify = AMPLIFY(config, **kwargs)

        if config.layer_norm_before_last_layer:
            self.decoder = transformer_engine.pytorch.LayerNormLinear(
                config.hidden_size,
                config.vocab_size,
                config.norm_eps,
                params_dtype=config.torch_dtype,
                normalization="RMSNorm" if config.rms_norm else "LayerNorm",
            )

        else:
            self.decoder = transformer_engine.pytorch.Linear(
                config.hidden_size, config.vocab_size, params_dtype=config.torch_dtype
            )

    def forward(
        self,
        input_ids,
        attention_mask=None,
        output_hidden_states=False,
        output_attentions=False,
        labels=None,
        **kwargs,
    ):
        outputs = self.amplify(
            input_ids,
            attention_mask,
            output_hidden_states,
            output_attentions,
            labels,
            **kwargs,
        )

        # Classification head with layer norm
        logits = self.decoder(outputs.last_hidden_state)

        if labels is not None:
            loss = nn.functional.cross_entropy(
                logits.view(-1, logits.size(-1)), labels.view(-1)
            )

        else:
            loss = None

        # Return logits or the output of the last hidden layer
        return MaskedLMOutput(
            loss=loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
        )