train_GeDi.py

# Adapted from https://github.com/huggingface/transformers/blob/21da895013a95e60df645b7d6b95f4a38f604759/examples/run_glue.py
# for training GPT-2 medium for sequence classification with GeDi objective


# coding=utf-8
# Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team.
# Copyright (c) 2018, 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 argparse
import glob
import json
import logging
import os
import random

import numpy as np
import torch
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from torch.utils.data.distributed import DistributedSampler
from tqdm import tqdm, trange
import sys


from modeling_gpt2 import GPT2LMHeadModel

from transformers import (
    WEIGHTS_NAME,
    AdamW,
    AlbertConfig,
    AlbertForSequenceClassification,
    AlbertTokenizer,
    BertConfig,
    BertForSequenceClassification,
    BertTokenizer,
    DistilBertConfig,
    DistilBertForSequenceClassification,
    DistilBertTokenizer,
    FlaubertConfig,
    FlaubertForSequenceClassification,
    FlaubertTokenizer,
    RobertaConfig,
    RobertaForSequenceClassification,
    RobertaTokenizer,
    XLMConfig,
    XLMForSequenceClassification,
    XLMRobertaConfig,
    XLMRobertaForSequenceClassification,
    XLMRobertaTokenizer,
    XLMTokenizer,
    XLNetConfig,
    XLNetForSequenceClassification,
    XLNetTokenizer,
    get_linear_schedule_with_warmup,
    GPT2Config,
    GPT2Tokenizer,
)
from transformers import glue_compute_metrics as compute_metrics
from transformers import glue_convert_examples_to_features as convert_examples_to_features
from transformers import glue_output_modes as output_modes
from transformers import glue_processors as processors

# https://github.com/huggingface/transformers/blob/master/src/transformers/data/metrics/__init__.py
def acc_and_f1(preds, labels):
    assert len(preds) == len(labels)
    acc = simple_accuracy(preds, labels)
    f1 = f1_score(y_true=labels, y_pred=preds)
    return {
        "acc": acc,
        "f1": f1,
        "acc_and_f1": (acc + f1) / 2,
    }
from sklearn.metrics import matthews_corrcoef, f1_score
def simple_accuracy(preds, labels):
        return (preds == labels).mean()

try:
    from torch.utils.tensorboard import SummaryWriter
except ImportError:
    from tensorboardX import SummaryWriter


logger = logging.getLogger(__name__)

ALL_MODELS = sum(
    (
        tuple(conf.pretrained_config_archive_map.keys())
        for conf in (
            BertConfig,
            XLNetConfig,
            XLMConfig,
            RobertaConfig,
            DistilBertConfig,
            AlbertConfig,
            XLMRobertaConfig,
            FlaubertConfig,
            GPT2Config
        )
    ),
    (),
)

MODEL_CLASSES = {
    "bert": (BertConfig, BertForSequenceClassification, BertTokenizer),
    "xlnet": (XLNetConfig, XLNetForSequenceClassification, XLNetTokenizer),
    "xlm": (XLMConfig, XLMForSequenceClassification, XLMTokenizer),
    "roberta": (RobertaConfig, RobertaForSequenceClassification, RobertaTokenizer),
    "distilbert": (DistilBertConfig, DistilBertForSequenceClassification, DistilBertTokenizer),
    "albert": (AlbertConfig, AlbertForSequenceClassification, AlbertTokenizer),
    "xlmroberta": (XLMRobertaConfig, XLMRobertaForSequenceClassification, XLMRobertaTokenizer),
    "flaubert": (FlaubertConfig, FlaubertForSequenceClassification, FlaubertTokenizer),
    "gpt2": (GPT2Config, GPT2LMHeadModel, GPT2Tokenizer)
}

#GPT2 added as per - https://huggingface.co/transformers/model_doc/gpt2.html
def add_sep(batch, sep_id):

    batch[0]
    len_list = (batch[1].sum(dim=1) - batch[2].sum(dim=1)).tolist()

    left_chunk = [x[:len_] for x,len_ in zip(batch[0],len_list)]
    right_chunk= [x[len_:] for x,len_ in zip(batch[0],len_list)]



    mid_chunk = [torch.Tensor(sep_id).type_as(x) for x in batch[0]]


    tensor_list = [torch.cat((left,mid,right)) for (left,mid,right) in
                   zip(left_chunk, mid_chunk, right_chunk)]
    return torch.stack(tensor_list)[:,:-1]

def set_seed(args):
    random.seed(args.seed)
    np.random.seed(args.seed)
    torch.manual_seed(args.seed)
    if args.n_gpu > 0:
        torch.cuda.manual_seed_all(args.seed)


def train(args, train_dataset, model, tokenizer):
    """ Train the model """
    if args.local_rank in [-1, 0]:
        tb_writer = SummaryWriter()

    args.train_batch_size = args.per_gpu_train_batch_size * max(1, args.n_gpu)
    train_sampler = RandomSampler(train_dataset) if args.local_rank == -1 else DistributedSampler(train_dataset)
    train_dataloader = DataLoader(train_dataset, sampler=train_sampler, batch_size=args.train_batch_size)

    if args.max_steps > 0:
        t_total = args.max_steps
        args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
    else:
        t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs

    # Prepare optimizer and schedule (linear warmup and decay)
    no_decay = ["bias", "LayerNorm.weight"]
    optimizer_grouped_parameters = [
        {
            "params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
            "weight_decay": args.weight_decay,
        },
        {"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], "weight_decay": 0.0},
    ]

    optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
    scheduler = get_linear_schedule_with_warmup(
        optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
    )


    # Check if saved optimizer or scheduler states exist
    if os.path.isfile(os.path.join(args.model_name_or_path, "optimizer.pt")) and os.path.isfile(
        os.path.join(args.model_name_or_path, "scheduler.pt")
    ):
        # Load in optimizer and scheduler states
        optimizer.load_state_dict(torch.load(os.path.join(args.model_name_or_path, "optimizer.pt")))
        scheduler.load_state_dict(torch.load(os.path.join(args.model_name_or_path, "scheduler.pt")))

    if args.fp16:
        try:
            from apex import amp
        except ImportError:
            raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
        model, optimizer = amp.initialize(model, optimizer, opt_level=args.fp16_opt_level)
        torch.cuda.empty_cache()

    # multi-gpu training (should be after apex fp16 initialization)
    if args.n_gpu > 1:
        model = torch.nn.DataParallel(model)

    # Distributed training (should be after apex fp16 initialization)
    if args.local_rank != -1:
        model = torch.nn.parallel.DistributedDataParallel(
            model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=False,
        )

    # Train!
    logger.info("***** Running training *****")
    logger.info("  Num examples = %d", len(train_dataset))
    logger.info("  Num Epochs = %d", args.num_train_epochs)
    logger.info("  Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
    logger.info(
        "  Total train batch size (w. parallel, distributed & accumulation) = %d",
        args.train_batch_size
        * args.gradient_accumulation_steps
        * (torch.distributed.get_world_size() if args.local_rank != -1 else 1),
    )
    logger.info("  Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
    logger.info("  Total optimization steps = %d", t_total)

    global_step = 0
    epochs_trained = 0
    steps_trained_in_current_epoch = 0
    # Check if continuing training from a checkpoint
    if os.path.exists(args.model_name_or_path):
        # set global_step to gobal_step of last saved checkpoint from model path
        global_step = int(args.model_name_or_path.split("-")[-1].split("/")[0])
        epochs_trained = global_step // (len(train_dataloader) // args.gradient_accumulation_steps)
        steps_trained_in_current_epoch = global_step % (len(train_dataloader) // args.gradient_accumulation_steps)

        logger.info("  Continuing training from checkpoint, will skip to saved global_step")
        logger.info("  Continuing training from epoch %d", epochs_trained)
        logger.info("  Continuing training from global step %d", global_step)
        logger.info("  Will skip the first %d steps in the first epoch", steps_trained_in_current_epoch)

    tr_loss, logging_loss = 0.0, 0.0
    model.zero_grad()
    train_iterator = trange(
        epochs_trained, int(args.num_train_epochs), desc="Epoch", disable=args.local_rank not in [-1, 0],
    )
    set_seed(args)  # Added here for reproductibility



    pt_id = tokenizer.encode(args.code_0)[0]
    nt_id = tokenizer.encode(args.code_1)[0]

    if args.threeway:
        nl_id = tokenizer.encode('neutral')[0]



    for epoch_ in train_iterator:

        epoch_iterator = train_dataloader
        for step, batch in enumerate(epoch_iterator):

            if args.sst5:
                split_lookup = {0:0, 1:1, 2:2, 3:3, 4:3, 5:3} #epoch:threshold_to_split
                threshold = split_lookup[epoch_]
                new_labels = ( batch[3] > threshold ).type_as(batch[3])
                batch[3] = new_labels


            # Skip past any already trained steps if resuming training
            if steps_trained_in_current_epoch > 0:
                steps_trained_in_current_epoch -= 1
                continue
            model.train()
            batch = tuple(t.to(args.device) for t in batch)
            if args.model_type == 'gpt2':

                batch_0 = add_sep(batch, tokenizer.encode('<sep>')) if args.add_sep else batch[0]

                #prepending tokens corresponding to 'positive' and 'negative' to the inputs
                seq_a = (torch.ones(batch_0.shape[0])*pt_id).type_as(batch_0).view(-1,1)
                seq_b = (torch.ones(batch_0.shape[0])*nt_id).type_as(batch_0).view(-1,1)

                if args.threeway:
                    seq_c = (torch.ones(batch_0.shape[0])*nl_id).type_as(batch_0).view(-1,1)

                seq_a = torch.cat((seq_a, batch_0), dim=1)[:,:-1]
                seq_b = torch.cat((seq_b, batch_0), dim=1)[:,:-1]
                bsz = seq_a.shape[0]
                if args.threeway:
                    seq_c = torch.cat((seq_c, batch_0), dim=1)[:,:-1]
                    seq_batched = torch.cat((seq_a,seq_b,seq_c),dim=0)

                else:
                    seq_batched = torch.cat((seq_a,seq_b),dim=0)
                #want to compute LM loss here so feeding inputs as labels
                inputs = {"input_ids": seq_batched, "attention_mask": None, "labels": seq_batched}

            else:
                assert args.model_type == 'gpt2' #let's not support other models for now
                inputs = {"input_ids": batch[0], "attention_mask": batch[1], "labels": batch[3]}

            outputs = model(**inputs) #modeling_gpt2.py modified to have none reduction
            losses = outputs[0].view(seq_batched.shape[0], -1)
            #loss mask includes first padded token
            if args.mask_eos_token:
                loss_mask = batch[1][:,:-1].to(torch.float16).cuda()
                if args.add_sep:
                    raise NotImplementedError
            else:
                loss_mask = batch[1][:,:-1].to(torch.float32).cuda()
                #appending with ones to account for the control code token being added
                if args.add_sep:
                    #adding the sep token would require extending the loss mask of ones by one position to the right (equivalent to prepending one on the left)
                    left_ = torch.ones(loss_mask.shape[0],2).type_as(loss_mask)
                    loss_mask = torch.cat((left_, loss_mask[:,:-2]), dim=1)
                else:
                    left_ = torch.ones(loss_mask.shape[0],1).type_as(loss_mask)
                    loss_mask = torch.cat((left_, loss_mask[:,:-1]), dim=1)


            loss_lengths = torch.sum(loss_mask,1,keepdim=True)

            if args.threeway:
                loss_a,loss_b,loss_c=torch.split(losses, bsz, dim=0)
                loss_a*=loss_mask
                loss_b*=loss_mask
                loss_c*=loss_mask

                #weights all batches equally even if sum of seqlens vary accross training iterations
                if False:
                    onehot_a = (batch[3]==0).to(torch.float16)
                    onehot_a = (batch[3]==0).to(torch.float16)
                    gen_loss_a = (batch[3]==0).to(torch.float16).unsqueeze(1)*loss_a/loss_lengths
                    gen_loss_b = (batch[3]==1).to(torch.float16).unsqueeze(1)*loss_b/loss_lengths
                    gen_loss_c = (batch[3]==2).to(torch.float16).unsqueeze(1)*loss_c/loss_lengths
                else:
                    onehot_a = (batch[3]==0).to(torch.float32)
                    onehot_a = (batch[3]==0).to(torch.float32)
                    gen_loss_a = (batch[3]==0).to(torch.float32).unsqueeze(1)*loss_a/loss_lengths
                    gen_loss_b = (batch[3]==1).to(torch.float32).unsqueeze(1)*loss_b/loss_lengths
                    gen_loss_c = (batch[3]==2).to(torch.float32).unsqueeze(1)*loss_c/loss_lengths

                gen_loss = torch.sum(gen_loss_a+gen_loss_b+gen_loss_c)/bsz
                if args.sum_loss:
                    loss_a = loss_a.sum(dim=1)
                    loss_b= loss_b.sum(dim=1)
                    loss_c= loss_c.sum(dim=1)

                else:
                    loss_a = (loss_a/loss_lengths).sum(dim=1)
                    loss_b= (loss_b/loss_lengths).sum(dim=1)
                    loss_c= (loss_c/loss_lengths).sum(dim=1)

            else:
                loss_a,loss_b=torch.split(losses, bsz, dim=0)

                loss_a*=loss_mask
                loss_b*=loss_mask
                if False:
                    gen_loss_a = (batch[3]==0).to(torch.float16).unsqueeze(1)*loss_a/loss_lengths
                    gen_loss_b = (batch[3]==1).to(torch.float16).unsqueeze(1)*loss_b/loss_lengths
                else:
                    gen_loss_a = (batch[3]==0).to(torch.float32).unsqueeze(1)*loss_a/loss_lengths
                    gen_loss_b = (batch[3]==1).to(torch.float32).unsqueeze(1)*loss_b/loss_lengths

                if args.jigsaw and args.jigsaw_no_toxic_gen:
                    gen_loss = torch.sum(gen_loss_a+gen_loss_b)/bsz

                else:
                    gen_loss = torch.sum(gen_loss_a+gen_loss_b)/bsz

                if args.sum_loss:
                    loss_a = loss_a.sum(dim=1)
                    loss_b= loss_b.sum(dim=1)

                else:
                    loss_a = (loss_a/loss_lengths).sum(dim=1)
                    loss_b= (loss_b/loss_lengths).sum(dim=1)

            if args.threeway:
                class_logits = torch.stack((-loss_a, -loss_b, -loss_c), dim=1)
            else:
                class_logits = torch.stack((-loss_a, -loss_b), dim=1) #(bsz, 2) dimensional
                batch[3][batch[3] == 2] = 1  #turning 3-ary to binary
            class_labels = batch[3]

            if args.logit_scale:
                if args.fp16:
                    if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                        class_logits*=model.logit_scale.float()
                    else:
                        class_logits*=model.module.logit_scale.float()
                else:
                    if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                        class_logits*=model.logit_scale
                    else:
                        class_logits*=model.module.logit_scale

            if args.outbias:
                if args.fp16:
                    if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                        class_logits+=model.bias.float()
                    else:
                        class_logits+= model.module.bias.float()
                else:
                    if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                        class_logits+=model.bias
                    else:
                        class_logits+=model.module.bias


            loss_fn = torch.nn.CrossEntropyLoss()
            loss = loss_fn(class_logits, class_labels)*args.disc_weight + args.gen_weight*gen_loss


            if np.isnan(loss.detach().cpu().numpy()):
                import pdb; pdb.set_trace()

            if args.n_gpu > 1:
                loss = loss.mean()  # mean() to average on multi-gpu parallel training
            if args.gradient_accumulation_steps > 1:
                loss = loss / args.gradient_accumulation_steps

            if args.fp16:
                with amp.scale_loss(loss, optimizer) as scaled_loss:
                    scaled_loss.backward()
            else:
                loss.backward()

            tr_loss += loss.item()

            if (step + 1) % args.gradient_accumulation_steps == 0:
                if args.fp16:
                    torch.nn.utils.clip_grad_norm_(amp.master_params(optimizer), args.max_grad_norm)
                else:
                    torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)

                optimizer.step()
                scheduler.step()  # Update learning rate schedule
                model.zero_grad()
                global_step += 1

                if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
                    logs = {}
                    if (
                        args.local_rank == -1 and args.evaluate_during_training
                    ):  # Only evaluate when single GPU otherwise metrics may not average well
                        results = evaluate(args, model, tokenizer)
                        for key, value in results.items():
                            eval_key = "eval_{}".format(key)
                            logs[eval_key] = value

                    loss_scalar = (tr_loss - logging_loss) / args.logging_steps
                    learning_rate_scalar = scheduler.get_lr()[0]
                    logs["learning_rate"] = learning_rate_scalar
                    logs["loss"] = loss_scalar
                    logging_loss = tr_loss

                    for key, value in logs.items():
                        tb_writer.add_scalar(key, value, global_step)
                    print(json.dumps({**logs, **{"step": global_step}}))

                if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
                    # Save model checkpoint
                    output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
                    if not os.path.exists(output_dir):
                        os.makedirs(output_dir)
                    model_to_save = (
                        model.module if hasattr(model, "module") else model
                    )  # Take care of distributed/parallel training
                    model_to_save.save_pretrained(output_dir)
                    tokenizer.save_pretrained(output_dir)

                    torch.save(args, os.path.join(output_dir, "training_args.bin"))
                    logger.info("Saving model checkpoint to %s", output_dir)

                    torch.save(optimizer.state_dict(), os.path.join(output_dir, "optimizer.pt"))
                    torch.save(scheduler.state_dict(), os.path.join(output_dir, "scheduler.pt"))
                    logger.info("Saving optimizer and scheduler states to %s", output_dir)

            if args.max_steps > 0 and global_step > args.max_steps:
                epoch_iterator.close()
                break
        if args.max_steps > 0 and global_step > args.max_steps:
            train_iterator.close()
            break

    if args.local_rank in [-1, 0]:
        tb_writer.close()

    return global_step, tr_loss / global_step


def evaluate(args, model, tokenizer, prefix=""):
    # Loop to handle MNLI double evaluation (matched, mis-matched)
    eval_task_names = ("mnli", "mnli-mm") if args.task_name == "mnli" else (args.task_name,)
    eval_outputs_dirs = (args.output_dir, args.output_dir + "-MM") if args.task_name == "mnli" else (args.output_dir,)

    results = {}
    for eval_task, eval_output_dir in zip(eval_task_names, eval_outputs_dirs):
        eval_dataset = load_and_cache_examples(args, eval_task, tokenizer, evaluate=True)

        if not os.path.exists(eval_output_dir) and args.local_rank in [-1, 0]:
            os.makedirs(eval_output_dir)

        args.eval_batch_size = args.per_gpu_eval_batch_size * max(1, args.n_gpu)
        # Note that DistributedSampler samples randomly
        eval_sampler = SequentialSampler(eval_dataset)
        eval_dataloader = DataLoader(eval_dataset, sampler=eval_sampler, batch_size=args.eval_batch_size)

        # multi-gpu eval
        if args.n_gpu > 1 and not isinstance(model, torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
            model = torch.nn.DataParallel(model)

        # Eval!
        logger.info("***** Running evaluation {} *****".format(prefix))
        logger.info("  Num examples = %d", len(eval_dataset))
        logger.info("  Batch size = %d", args.eval_batch_size)
        eval_loss = 0.0
        overall_gen_loss = 0.0
        nb_eval_steps = 0
        preds = None
        out_label_ids = None



        for batch in eval_dataloader:
            model.eval()
            batch = tuple(t.to(args.device) for t in batch)

            with torch.no_grad():

                batch_0 = add_sep(batch, tokenizer.encode('<sep>')) if args.add_sep else batch[0]

                #this is assuming you get only one token for the words 'positive' and 'negative'
                pt_id = tokenizer.encode(args.code_0)[0] #pt_id = tokenizer.encode('positive')[0]
                nt_id = tokenizer.encode(args.code_1)[0] #nt_id = tokenizer.encode('negative')[0]

                if args.threeway:
                    nl_id = tokenizer.encode('neutral')[0]

                #prepending tokens corresponding to 'positive' and 'negative' to the inputs
                seq_a = (torch.ones(batch_0.shape[0])*pt_id).type_as(batch_0).view(-1,1)
                seq_b = (torch.ones(batch_0.shape[0])*nt_id).type_as(batch_0).view(-1,1)

                if args.threeway:
                    seq_c = (torch.ones(batch_0.shape[0])*nl_id).type_as(batch_0).view(-1,1)

                seq_a = torch.cat((seq_a, batch_0), dim=1)[:,:-1]
                seq_b = torch.cat((seq_b, batch_0), dim=1)[:,:-1]
                bsz = seq_a.shape[0]
                if args.threeway:
                    seq_c = torch.cat((seq_c, batch_0), dim=1)[:,:-1]
                    seq_batched = torch.cat((seq_a,seq_b,seq_c),dim=0)

                else:
                    seq_batched = torch.cat((seq_a,seq_b),dim=0)

                #want to compute LM loss here so feeding inputs as labels
                inputs = {"input_ids": seq_batched, "attention_mask": None, "labels": seq_batched}

                outputs = model(**inputs) #modeling_gpt2.py changed to have none reduction
                losses = outputs[0].view(seq_batched.shape[0], -1)

                #loss mask includes first padded token
                if args.mask_eos_token:
                    loss_mask = batch[1][:,:-1].to(torch.float16).cuda()
                else:
                    loss_mask = batch[1][:,:-1].to(torch.float32).cuda()
                    #appending with ones to account for the control code token being added
                    left_ = torch.ones(loss_mask.shape[0],1).type_as(loss_mask)
                    loss_mask = torch.cat((left_, loss_mask[:,:-1]), dim=1)



                loss_lengths = torch.sum(loss_mask,1,keepdim=True)

                if args.threeway:
                    loss_a,loss_b,loss_c=torch.split(losses, bsz, dim=0)
                    loss_a*=loss_mask
                    loss_b*=loss_mask
                    loss_c*=loss_mask

                    #weights all batches equally even if sum of seqlens vary accross training iterations
                    if False:
                        gen_loss_a = (batch[3]==0).to(torch.float16).unsqueeze(1)*loss_a/loss_lengths
                        gen_loss_b = (batch[3]==1).to(torch.float16).unsqueeze(1)*loss_b/loss_lengths
                        gen_loss_c = (batch[3]==2).to(torch.float16).unsqueeze(1)*loss_c/loss_lengths
                    else:
                        gen_loss_a = (batch[3]==0).to(torch.float32).unsqueeze(1)*loss_a/loss_lengths
                        gen_loss_b = (batch[3]==1).to(torch.float32).unsqueeze(1)*loss_b/loss_lengths
                        gen_loss_c = (batch[3]==2).to(torch.float32).unsqueeze(1)*loss_c/loss_lengths
                    gen_loss = torch.sum(gen_loss_a+gen_loss_b+gen_loss_c)/bsz
                    if args.sum_loss:
                        loss_a = loss_a.sum(dim=1)
                        loss_b= loss_b.sum(dim=1)
                        loss_c= loss_c.sum(dim=1)

                    else:
                        loss_a = (loss_a/loss_lengths).sum(dim=1)
                        loss_b= (loss_b/loss_lengths).sum(dim=1)
                        loss_c= (loss_c/loss_lengths).sum(dim=1)

                else:
                    loss_a,loss_b=torch.split(losses, bsz, dim=0)

                    loss_a*=loss_mask
                    loss_b*=loss_mask

                    #weights all batches equally even if sum of seqlens vary accross training iterations
                    if False:
                        gen_loss_a = (batch[3]==0).to(torch.float16).unsqueeze(1)*loss_a/loss_lengths
                        gen_loss_b = (batch[3]==1).to(torch.float16).unsqueeze(1)*loss_b/loss_lengths
                    else:
                        gen_loss_a = (batch[3]==0).to(torch.float32).unsqueeze(1)*loss_a/loss_lengths
                        gen_loss_b = (batch[3]==1).to(torch.float32).unsqueeze(1)*loss_b/loss_lengths

                    gen_loss = torch.sum(gen_loss_a+gen_loss_b)/bsz

                    if args.sum_loss:
                        loss_a = loss_a.sum(dim=1)
                        loss_b= loss_b.sum(dim=1)

                    else:
                        loss_a = (loss_a/loss_lengths).sum(dim=1)
                        loss_b= (loss_b/loss_lengths).sum(dim=1)

                if args.threeway:
                    class_logits = torch.stack((-loss_a, -loss_b, -loss_c), dim=1)
                else:
                    class_logits = torch.stack((-loss_a, -loss_b), dim=1) #(bsz, 2) dimensional
                    batch[3][batch[3] == 2] = 1  #turning 3-ary to binary





                class_labels = batch[3]


                loss_fn = torch.nn.CrossEntropyLoss()

                if args.logit_scale:
                    if args.fp16:
                        if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                            class_logits*=model.logit_scale.float()
                        else:
                            class_logits*=model.module.logit_scale.float()
                    else:
                        if not isinstance(model,torch.nn.DataParallel) and not isinstance(model,torch.nn.parallel.DistributedDataParallel):
                            class_logits*=model.logit_scale
                        else:
                            class_logits*=model.module.logit_scale.float()

                if args.outbias:
                    if args.fp16:
                        class_logits+=model.bias.float()
                    else:
                        class_logits+=model.bias


                loss = loss_fn(class_logits, class_labels)

                tmp_eval_loss = loss
                tmp_gen_loss = gen_loss
                logits = class_logits


                eval_loss += tmp_eval_loss.mean().item()
                overall_gen_loss += tmp_gen_loss.mean().item()
            nb_eval_steps += 1
            if preds is None:
                preds = logits.detach().cpu().numpy()
                out_label_ids = class_labels.detach().cpu().numpy()
            else:
                preds = np.append(preds, logits.detach().cpu().numpy(), axis=0)
                out_label_ids = np.append(out_label_ids, class_labels.detach().cpu().numpy(), axis=0)


        eval_loss = eval_loss / nb_eval_steps
        overall_gen_loss = overall_gen_loss / nb_eval_steps
        if args.output_mode == "classification":
            preds = np.argmax(preds, axis=1)
        elif args.output_mode == "regression":
            preds = np.squeeze(preds)
        result = compute_metrics(eval_task, preds, out_label_ids)

        results.update(result)

        #avg gen_loss over all val samples
        #will be used to compute perplexity for the conditional seq do_lower_case
        result.update({'overall_gen_loss':overall_gen_loss})

        output_eval_file = os.path.join(eval_output_dir, prefix, "eval_results.txt")
        print(result)
        print(results)
        with open(output_eval_file, "w") as writer:
            logger.info("***** Eval results {} *****".format(prefix))
            for key in sorted(result.keys()):
                logger.info("  %s = %s", key, str(result[key]))
                writer.write("%s = %s\n" % (key, str(result[key])))
            if eval_task == 'cola':
                accf1 = acc_and_f1(preds, out_label_ids)
                print(accf1)
                for key in sorted(accf1.keys()):
                    logger.info("  %s = %s", key, str(accf1[key]))
                    writer.write("%s = %s\n" % (key, str(accf1[key])))

    return results


def load_and_cache_examples(args, task, tokenizer, evaluate=False):

    if args.local_rank not in [-1, 0] and not evaluate:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    processor = processors[task]()
    output_mode = output_modes[task]
    # Load data features from cache or dataset file
    cached_features_file = os.path.join(
        args.data_dir,
        "cached_{}_{}_{}_{}".format(
            "dev" if evaluate else "train",
            list(filter(None, args.model_name_or_path.split("/"))).pop(),
            str(args.max_seq_length),
            str(task),
        ),
    )
    if os.path.exists(cached_features_file) and not args.overwrite_cache:
        logger.info("Loading features from cached file %s", cached_features_file)
        features = torch.load(cached_features_file)
    else:
        logger.info("Creating features from dataset file at %s", args.data_dir)
        label_list = processor.get_labels()
        if task in ["mnli", "mnli-mm"] and args.model_type in ["roberta", "xlmroberta"]:
            # HACK(label indices are swapped in RoBERTa pretrained model)
            label_list[1], label_list[2] = label_list[2], label_list[1]
        examples = (
            processor.get_dev_examples(args.data_dir) if evaluate else processor.get_train_examples(args.data_dir)
        )

        if args.model_type == 'gpt2': #setting pad token for GPT-2
            tokenizer.pad_token = '[PAD]'

        if args.sst5:
            label_list = ['0','1','2','3','4']

        features = convert_examples_to_features(
            examples,
            tokenizer,
            label_list=label_list,
            max_length=args.max_seq_length,
            output_mode=output_mode,
            pad_on_left=bool(args.model_type in ["xlnet"]),  # pad on the left for xlnet
            pad_token=tokenizer.convert_tokens_to_ids([tokenizer.pad_token])[0],
            pad_token_segment_id=4 if args.model_type in ["xlnet"] else 0,
        )
        if args.local_rank in [-1, 0]:
            logger.info("Saving features into cached file %s", cached_features_file)
            torch.save(features, cached_features_file)

    if args.local_rank == 0 and not evaluate:
        torch.distributed.barrier()  # Make sure only the first process in distributed training process the dataset, and the others will use the cache

    # Convert to Tensors and build dataset
    all_input_ids = torch.tensor([f.input_ids for f in features], dtype=torch.long)
    all_attention_mask = torch.tensor([f.attention_mask for f in features], dtype=torch.long)
    all_token_type_ids = torch.tensor([f.token_type_ids for f in features], dtype=torch.long)
    if output_mode == "classification":
        all_labels = torch.tensor([f.label for f in features], dtype=torch.long)
    elif output_mode == "regression":
        all_labels = torch.tensor([f.label for f in features], dtype=torch.float)

    dataset = TensorDataset(all_input_ids, all_attention_mask, all_token_type_ids, all_labels)
    return dataset


def main():
    parser = argparse.ArgumentParser()

    # Required parameters
    parser.add_argument(
        "--data_dir",
        default=None,
        type=str,
        required=True,
        help="The input data dir. Should contain the .tsv files (or other data files) for the task.",
    )
    parser.add_argument(
        "--model_type",
        default=None,
        type=str,
        required=True,
        help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
    )
    parser.add_argument(
        "--model_name_or_path",
        default=None,
        type=str,
        required=True,
        help="Path to pre-trained model or shortcut name selected in the list: " + ", ".join(ALL_MODELS),
    )
    parser.add_argument(
        "--task_name",
        default=None,
        type=str,
        required=True,
        help="The name of the task to train selected in the list: " + ", ".join(processors.keys()),
    )
    parser.add_argument(
        "--output_dir",
        default=None,
        type=str,
        required=True,
        help="The output directory where the model predictions and checkpoints will be written.",
    )

    #new generative classifier specific parameters
    parser.add_argument("--dropout",default=0.1,type=float, help="dropout prob")
    parser.add_argument("--gen_weight",default=0.0,type=float, help="scalar multiple for generative loss (lambda)")

    parser.add_argument("--logit_scale",action="store_true",help="learns to scale logits for classification")
    parser.add_argument("--threeway", action="store_true", help="does 3-way classification")
    parser.add_argument("--sum_loss",action="store_true", help="sums losses")
    parser.add_argument("--outbias",action="store_true", help="learns output bias for each class")

    # Other parameters
    parser.add_argument(
        "--config_name", default="", type=str, help="Pretrained config name or path if not the same as model_name",
    )
    parser.add_argument(
        "--tokenizer_name",
        default="",
        type=str,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--cache_dir",
        default="",
        type=str,
        help="Where do you want to store the pre-trained models downloaded from s3",
    )
    parser.add_argument(
        "--max_seq_length",
        default=128,
        type=int,
        help="The maximum total input sequence length after tokenization. Sequences longer "
        "than this will be truncated, sequences shorter will be padded.",
    )
    parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
    parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")




    parser.add_argument(
        "--evaluate_during_training", action="store_true", help="Run evaluation during training at each logging step.",
    )
    parser.add_argument(
        "--do_lower_case", action="store_true", help="Set this flag if you are using an uncased model.",
    )

    parser.add_argument(
        "--per_gpu_train_batch_size", default=8, type=int, help="Batch size per GPU/CPU for training.",
    )
    parser.add_argument(
        "--per_gpu_eval_batch_size", default=8, type=int, help="Batch size per GPU/CPU for evaluation.",
    )
    parser.add_argument(
        "--gradient_accumulation_steps",
        type=int,
        default=1,
        help="Number of updates steps to accumulate before performing a backward/update pass.",
    )
    parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
    parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight decay if we apply some.")
    parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
    parser.add_argument(
        "--num_train_epochs", default=3.0, type=float, help="Total number of training epochs to perform.",
    )
    parser.add_argument(
        "--max_steps",
        default=-1,
        type=int,
        help="If > 0: set total number of training steps to perform. Override num_train_epochs.",
    )
    parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")

    parser.add_argument("--logging_steps", type=int, default=500, help="Log every X updates steps.")
    parser.add_argument("--save_steps", type=int, default=500, help="Save checkpoint every X updates steps.")
    parser.add_argument(
        "--eval_all_checkpoints",
        action="store_true",
        help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number",
    )
    parser.add_argument("--no_cuda", action="store_true", help="Avoid using CUDA when available")
    parser.add_argument(
        "--overwrite_output_dir", action="store_true", help="Overwrite the content of the output directory",
    )
    parser.add_argument(
        "--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets",
    )
    parser.add_argument("--seed", type=int, default=42, help="random seed for initialization")

    parser.add_argument(
        "--fp16",
        action="store_true",
        help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
    )
    parser.add_argument(
        "--fp16_opt_level",
        type=str,
        default="O1",
        help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
        "See details at https://nvidia.github.io/apex/amp.html",
    )
    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
    parser.add_argument("--server_ip", type=str, default="", help="For distant debugging.")
    parser.add_argument("--server_port", type=str, default="", help="For distant debugging.")
    parser.add_argument("--mask_eos_token", action="store_true",
        help="whether to mask eos token loss or not; prefer masking if training for DA",
    )
    parser.add_argument("--add_sep", action="store_true",
                        help="Include sep token if this arg is used between the two sentences in a pair | can/should be used for mrpc/mnli/qqp/qnli")
    parser.add_argument("--sst5", action="store_true",
                        help="custom ops for SST-5")
    parser.add_argument("--jigsaw", action="store_true", help="custom setup for jigsaw")
    parser.add_argument("--jigsaw_no_toxic_gen", action="store_true", help="custom setup for jigsaw - gen_loss used only for non-toxic samples | check training loop")
    parser.add_argument("--code_0", type=str, default="negative", help="control code to be used for code 1 of 2 (we support 3 at most - with the third one = 'neutral' for now)")
    parser.add_argument("--code_1", type=str, default="positive", help="control code to be used for code 2 of 2 (we support 3 at most - with the third one = 'neutral' for now)")

#     args = parser.parse_args()
    args, unknown = parser.parse_known_args()
    if (
        os.path.exists(args.output_dir)
        and os.listdir(args.output_dir)
        and args.do_train
        and not args.overwrite_output_dir
    ):
        raise ValueError(
            "Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(
                args.output_dir
            )
        )

    # Setup distant debugging if needed
    if args.server_ip and args.server_port:
        # Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
        import ptvsd

        print("Waiting for debugger attach")
        ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
        ptvsd.wait_for_attach()

    # Setup CUDA, GPU & distributed training
    if args.local_rank == -1 or args.no_cuda:
        device = torch.device("cuda" if torch.cuda.is_available() and not args.no_cuda else "cpu")
        args.n_gpu = torch.cuda.device_count()
    else:  # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
        torch.cuda.set_device(args.local_rank)
        device = torch.device("cuda", args.local_rank)
        torch.distributed.init_process_group(backend="nccl")
        args.n_gpu = 1
    args.device = device
    args.disc_weight = 1.0 - args.gen_weight
    # Setup logging
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s -   %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN,
    )
    logger.warning(
        "Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
        args.local_rank,
        device,
        args.n_gpu,
        bool(args.local_rank != -1),
        args.fp16,
    )

    # Set seed
    set_seed(args)


    # Prepare GLUE task
    args.task_name = args.task_name.lower()
    if args.task_name not in processors:
        raise ValueError("Task not found: %s" % (args.task_name))
    processor = processors[args.task_name]()
    args.output_mode = output_modes[args.task_name]
    label_list = processor.get_labels()
    num_labels = len(label_list)

    # Load pretrained model and tokenizer
    if args.local_rank not in [-1, 0]:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab

    args.model_type = args.model_type.lower()
    config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
    config = config_class.from_pretrained(
        args.config_name if args.config_name else args.model_name_or_path,
        num_labels=num_labels,
        finetuning_task=args.task_name,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
    if args.outbias:
        if args.threeway:
            config.nbias=3
        else:
            config.nbias=2
    else:
        config.nbias=0

    config.embd_pdrop = args.dropout
    config.attn_pdrop = args.dropout
    config.resid_pdrop = args.dropout
    if args.logit_scale:
        config.logit_scale=True
    else:
        config.logit_scale=False

    tokenizer = tokenizer_class.from_pretrained(
        args.tokenizer_name if args.tokenizer_name else args.model_name_or_path,
        do_lower_case=args.do_lower_case,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
    if args.add_sep:
        special_tokens_dict = {'sep_token': '<SEP>'}
        tokenizer.add_special_tokens(special_tokens_dict)
    config.output_past = True #https://github.com/huggingface/transformers/pull/3734
    model = model_class.from_pretrained(
        args.model_name_or_path,
        from_tf=bool(".ckpt" in args.model_name_or_path),
        config=config,
        cache_dir=args.cache_dir if args.cache_dir else None,
    )
    model.resize_token_embeddings(len(tokenizer))

    if args.local_rank == 0:
        torch.distributed.barrier()  # Make sure only the first process in distributed training will download model & vocab


    model.to(args.device)




    logger.info("Training/evaluation parameters %s", args)

    # Training
    if args.do_train:
        train_dataset = load_and_cache_examples(args, args.task_name, tokenizer, evaluate=False)
        global_step, tr_loss = train(args, train_dataset, model, tokenizer)
        logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)

    # Saving best-practices: if you use defaults names for the model, you can reload it using from_pretrained()
    if args.do_train and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
        # Create output directory if needed
        if not os.path.exists(args.output_dir) and args.local_rank in [-1, 0]:
            os.makedirs(args.output_dir)

        logger.info("Saving model checkpoint to %s", args.output_dir)
        # Save a trained model, configuration and tokenizer using `save_pretrained()`.
        # They can then be reloaded using `from_pretrained()`
        model_to_save = (
            model.module if hasattr(model, "module") else model
        )  # Take care of distributed/parallel training
        model_to_save.save_pretrained(args.output_dir)
        tokenizer.save_pretrained(args.output_dir)

        # Good practice: save your training arguments together with the trained model
        torch.save(args, os.path.join(args.output_dir, "training_args.bin"))

        # Load a trained model and vocabulary that you have fine-tuned
        model = model_class.from_pretrained(args.output_dir)
        tokenizer = tokenizer_class.from_pretrained(args.output_dir)
        model.to(args.device)

    # Evaluation
    results = {}
    if args.do_eval and args.local_rank in [-1, 0]:
        tokenizer = tokenizer_class.from_pretrained(args.output_dir, do_lower_case=args.do_lower_case)
        checkpoints = [args.output_dir]
        if args.eval_all_checkpoints:
            checkpoints = list(
                os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + "/**/" + WEIGHTS_NAME, recursive=True))
            )
            logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN)  # Reduce logging
        logger.info("Evaluate the following checkpoints: %s", checkpoints)
        for checkpoint in checkpoints:
            global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
            prefix = checkpoint.split("/")[-1] if checkpoint.find("checkpoint") != -1 else ""

            model = model_class.from_pretrained(checkpoint)
            model.to(args.device)
            result = evaluate(args, model, tokenizer, prefix=prefix)
            result = dict((k + "_{}".format(global_step), v) for k, v in result.items())
            results.update(result)

    return results


if __name__ == "__main__":
    main()