initial commit

model/model.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 wrapping BertModel."""
+
+import torch
+
+from .modeling import BertConfig
+from .modeling import BertForPreTraining
+from .modeling import BertLayerNorm
+
+
+def get_params_for_weight_decay_optimization(module):
+
+    weight_decay_params = {'params': []}
+    no_weight_decay_params = {'params': [], 'weight_decay': 0}
+    for module_ in module.modules():
+        if isinstance(module_, (BertLayerNorm, torch.nn.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
+
+
+class BertModel(torch.nn.Module):
+
+    def __init__(self, tokenizer, args):
+        super(BertModel, self).__init__()
+        if args.pretrained_bert:
+            self.model = BertForPreTraining.from_pretrained(
+                args.tokenizer_model_type,
+                cache_dir=args.cache_dir,
+                fp32_layernorm=args.fp32_layernorm,
+                fp32_embedding=args.fp32_embedding,
+                layernorm_epsilon=args.layernorm_epsilon)
+        else:
+            if args.intermediate_size is None:
+                intermediate_size = 4 * args.hidden_size
+            else:
+                intermediate_size = args.intermediate_size
+            self.config = BertConfig(
+                tokenizer.num_tokens,
+                hidden_size=args.hidden_size,
+                num_hidden_layers=args.num_layers,
+                num_attention_heads=args.num_attention_heads,
+                intermediate_size=intermediate_size,
+                hidden_dropout_prob=args.hidden_dropout,
+                attention_probs_dropout_prob=args.attention_dropout,
+                max_position_embeddings=args.max_position_embeddings,
+                type_vocab_size=tokenizer.num_type_tokens,
+                fp32_layernorm=args.fp32_layernorm,
+                fp32_embedding=args.fp32_embedding,
+                fp32_tokentypes=args.fp32_tokentypes,
+                layernorm_epsilon=args.layernorm_epsilon)
+            self.model = BertForPreTraining(self.config)
+
+    def forward(self, input_tokens, token_type_ids=None,
+                attention_mask=None, checkpoint_activations=False):
+        return self.model(
+            input_tokens, token_type_ids, attention_mask,
+            checkpoint_activations=checkpoint_activations)
+
+    def state_dict(self, destination=None, prefix='', keep_vars=False):
+        return self.model.state_dict(destination=destination, prefix=prefix,
+                                     keep_vars=keep_vars)
+
+    def load_state_dict(self, state_dict, strict=True):
+        return self.model.load_state_dict(state_dict, strict=strict)

optim/__init__.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.
+
+from torch.optim import SGD
+from torch.optim import Adadelta
+from torch.optim import Adagrad
+from torch.optim import SparseAdam
+from torch.optim import Adamax
+from torch.optim import SGD
+from torch.optim import Rprop
+from torch.optim import RMSprop
+from torch.optim import Optimizer
+from torch.optim import LBFGS
+from .adam import Adam

optim/adam.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.
+
+import math
+import torch
+from torch.optim import Optimizer
+
+
+class Adam(Optimizer):
+    r"""Implements Adam algorithm.
+
+    It has been proposed in `Adam: A Method for Stochastic Optimization`_.
+
+    Arguments:
+        params (iterable): iterable of parameters to optimize or dicts defining
+            parameter groups
+        lr (float, optional): learning rate (default: 1e-3)
+        betas (Tuple[float, float], optional): coefficients used for computing
+            running averages of gradient and its square (default: (0.9, 0.999))
+        eps (float, optional): term added to the denominator to improve
+            numerical stability (default: 1e-8)
+        weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
+        amsgrad (boolean, optional): whether to use the AMSGrad variant of this
+            algorithm from the paper `On the Convergence of Adam and Beyond`_
+            (default: False)
+
+    .. _Adam\: A Method for Stochastic Optimization:
+        https://arxiv.org/abs/1412.6980
+    .. _On the Convergence of Adam and Beyond:
+        https://openreview.net/forum?id=ryQu7f-RZ
+    """
+
+    def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8,
+                 weight_decay=0, amsgrad=False):
+        if not 0.0 <= lr:
+            raise ValueError("Invalid learning rate: {}".format(lr))
+        if not 0.0 <= eps:
+            raise ValueError("Invalid epsilon value: {}".format(eps))
+        if not 0.0 <= betas[0] < 1.0:
+            raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
+        if not 0.0 <= betas[1] < 1.0:
+            raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
+        defaults = dict(lr=lr, betas=betas, eps=eps,
+                        weight_decay=weight_decay, amsgrad=amsgrad)
+        super(Adam, self).__init__(params, defaults)
+
+    def __setstate__(self, state):
+        super(Adam, self).__setstate__(state)
+        for group in self.param_groups:
+            group.setdefault('amsgrad', False)
+
+    def step(self, closure=None):
+        """Performs a single optimization step.
+
+        Arguments:
+            closure (callable, optional): A closure that reevaluates the model
+                and returns the loss.
+        """
+        loss = None
+        if closure is not None:
+            loss = closure()
+
+        for group in self.param_groups:
+            for p in group['params']:
+                if p.grad is None:
+                    continue
+                grad = p.grad.data
+                if grad.is_sparse:
+                    raise RuntimeError('Adam does not support sparse gradients, please consider SparseAdam instead')
+                amsgrad = group['amsgrad']
+
+                state = self.state[p]
+
+                # State initialization
+                if len(state) == 0:
+                    state['step'] = 0
+                    # Exponential moving average of gradient values
+                    state['exp_avg'] = torch.zeros_like(p.data)
+                    # Exponential moving average of squared gradient values
+                    state['exp_avg_sq'] = torch.zeros_like(p.data)
+                    if amsgrad:
+                        # Maintains max of all exp. moving avg. of sq. grad. values
+                        state['max_exp_avg_sq'] = torch.zeros_like(p.data)
+
+                exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq']
+                if amsgrad:
+                    max_exp_avg_sq = state['max_exp_avg_sq']
+                beta1, beta2 = group['betas']
+
+                state['step'] += 1
+
+
+                # Decay the first and second moment running average coefficient
+                exp_avg.mul_(beta1).add_(1 - beta1, grad)
+                exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)
+                if amsgrad:
+                    # Maintains the maximum of all 2nd moment running avg. till now
+                    torch.max(max_exp_avg_sq, exp_avg_sq, out=max_exp_avg_sq)
+                    # Use the max. for normalizing running avg. of gradient
+                    denom = max_exp_avg_sq.sqrt().add_(group['eps'])
+                else:
+                    denom = exp_avg_sq.sqrt().add_(group['eps'])
+
+                bias_correction1 = 1 - beta1 ** state['step']
+                bias_correction2 = 1 - beta2 ** state['step']
+                step_size = group['lr']# * math.sqrt(bias_correction2) / bias_correction1
+                if group['weight_decay'] != 0:
+                    p.data.add_(-step_size * group['weight_decay'], p.data)
+
+                p.data.addcdiv_(-step_size, exp_avg, denom)
+
+        return loss

requirements.txt

+nltk>=3.4
+numpy>=1.15.4
+pandas>=0.24.0
+sentencepiece>=0.1.8
+tensorflow>=1.12.0

scripts/pretrain_bert.sh

+#!/bin/bash
+
+RANK=0
+WORLD_SIZE=1
+
+python pretrain_bert.py \
+    --batch-size 4 \
+    --tokenizer-type BertWordPieceTokenizer \
+    --cache-dir cache_dir \
+    --tokenizer-model-type bert-large-uncased \
+    --vocab-size 30522 \
+    --train-data wikipedia \
+    --loose-json \
+    --text-key text \
+    --split 1000,1,1 \
+    --lazy-loader \
+    --max-preds-per-seq 80 \
+    --seq-length 512 \
+    --max-position-embeddings 512 \
+    --num-layers 24 \
+    --hidden-size 1024 \
+    --intermediate-size 4096 \
+    --num-attention-heads 16 \
+    --hidden-dropout 0.1 \
+    --attention-dropout 0.1 \
+    --train-iters 1000000 \
+    --lr 0.0001 \
+    --lr-decay-style linear \
+    --lr-decay-iters 990000 \
+    --warmup .01 \
+    --weight-decay 1e-2 \
+    --clip-grad 1.0 \
+    --fp16 \
+    --fp32-layernorm \
+    --fp32-embedding \
+    --hysteresis 2 \
+    --num-workers 2