decode.py

#   Copyright (c) 2018 PaddlePaddle Authors. 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 __future__ import division
from __future__ import absolute_import
from __future__ import print_function
from __future__ import unicode_literals

import sys
import io
import re
import argparse
import logging
import json
import numpy as np
from pathlib import Path
from collections import namedtuple

import paddle as P
from paddle.nn import functional as F

from ernie.modeling_ernie import ErnieModel, ErnieModelForPretraining, ErnieModelForGeneration
from ernie.modeling_ernie import _build_linear, _build_ln, append_name
from ernie.tokenizing_ernie import ErnieTokenizer

from propeller import log
import propeller.paddle as propeller


@np.vectorize
def rev_lookup(i):
    return rev_dict[i]


def gen_bias(encoder_inputs, decoder_inputs, step):
    decoder_bsz, decoder_seqlen = decoder_inputs.shape[:2]
    attn_bias = P.reshape(
        P.arange(
            0, decoder_seqlen, 1, dtype='float32') + 1, [1, -1, 1])
    decoder_bias = P.cast(
        (P.matmul(
            attn_bias, 1. / attn_bias, transpose_y=True) >= 1.),
        'float32')  #[1, 1, decoderlen, decoderlen]
    encoder_bias = P.unsqueeze(
        P.cast(P.ones_like(encoder_inputs), 'float32'),
        [1])  #[bsz, 1, encoderlen]
    encoder_bias = P.tile(
        encoder_bias, [1, decoder_seqlen, 1])  #[bsz,decoderlen, encoderlen]
    decoder_bias = P.tile(decoder_bias,
                          [decoder_bsz, 1, 1])  #[bsz, decoderlen, decoderlen]
    if step > 0:
        bias = P.concat([
            encoder_bias, P.ones([decoder_bsz, decoder_seqlen, step],
                                 'float32'), decoder_bias
        ], -1)
    else:
        bias = P.concat([encoder_bias, decoder_bias], -1)
    return bias


#def make_data(tokenizer, inputs, max_encode_len):
#    all_ids, all_sids = [], []
#    for i in inputs:
#        q_ids, q_sids = tokenizer.build_for_ernie(
#                np.array(
#                    tokenizer.convert_tokens_to_ids(i.split(' '))[: max_encode_len-2],
#                    dtype=np.int64
#                    )
#                )
#        all_ids.append(q_ids)
#        all_sids.append(q_sids)
#    ml = max(map(len, all_ids))
#    all_ids = [np.pad(i, [0, ml-len(i)], mode='constant')for i in all_ids]
#    all_sids = [np.pad(i, [0, ml-len(i)], mode='constant')for i in all_sids]
#    all_ids = np.stack(all_ids, 0)
#    all_sids = np.stack(all_sids, 0)
#    return all_ids, all_sids


def greedy_search_infilling(model,
                            q_ids,
                            q_sids,
                            sos_id,
                            eos_id,
                            attn_id,
                            max_encode_len=640,
                            max_decode_len=100,
                            tgt_type_id=3):
    model.eval()
    with P.no_grad():
        #log.debug(q_ids.numpy().tolist())
        _, logits, info = model(q_ids, q_sids)
        gen_ids = P.argmax(logits, -1)
        d_batch, d_seqlen = q_ids.shape
        seqlen = P.cast(q_ids != 0, 'int64').sum(1, keepdim=True)
        log.debug(seqlen.numpy())
        log.debug(d_seqlen)
        has_stopped = np.zeros([d_batch], dtype=np.bool)
        gen_seq_len = np.zeros([d_batch], dtype=np.int64)
        output_ids = []

        past_cache = info['caches']

        cls_ids = P.ones([d_batch], dtype='int64') * sos_id
        attn_ids = P.ones([d_batch], dtype='int64') * attn_id
        ids = P.stack([cls_ids, attn_ids], -1)
        for step in range(max_decode_len):
            log.debug('decode step %d' % step)
            bias = gen_bias(q_ids, ids, step)
            pos_ids = P.to_tensor(
                np.tile(
                    np.array(
                        [[step, step + 1]], dtype=np.int64), [d_batch, 1]))
            pos_ids += seqlen
            _, logits, info = model(
                ids,
                P.ones_like(ids) * tgt_type_id,
                pos_ids=pos_ids,
                attn_bias=bias,
                past_cache=past_cache)
            gen_ids = P.argmax(logits, -1)

            past_cached_k, past_cached_v = past_cache
            cached_k, cached_v = info['caches']
            cached_k = [
                P.concat([pk, k[:, :1, :]], 1)
                for pk, k in zip(past_cached_k, cached_k)
            ]  # concat cached
            cached_v = [
                P.concat([pv, v[:, :1, :]], 1)
                for pv, v in zip(past_cached_v, cached_v)
            ]
            past_cache = (cached_k, cached_v)

            gen_ids = gen_ids[:, 1]
            ids = P.stack([gen_ids, attn_ids], 1)

            gen_ids = gen_ids.numpy()
            has_stopped |= (gen_ids == eos_id).astype(np.bool)
            gen_seq_len += (1 - has_stopped.astype(np.int64))
            output_ids.append(gen_ids.tolist())
            if has_stopped.all():
                #log.debug('exit because all done')
                break
            #if step == 1: break
        output_ids = np.array(output_ids).transpose([1, 0])
    return output_ids


BeamSearchState = namedtuple('BeamSearchState',
                             ['log_probs', 'lengths', 'finished'])
BeamSearchOutput = namedtuple('BeamSearchOutput',
                              ['scores', 'predicted_ids', 'beam_parent_ids'])


def log_softmax(x):
    e_x = np.exp(x - np.max(x))
    return np.log(e_x / e_x.sum())


def mask_prob(p, onehot_eos, finished):
    is_finished = P.cast(P.reshape(finished, [-1, 1]) != 0, 'float32')
    p = is_finished * (1. - P.cast(onehot_eos, 'float32')) * -9999. + (
        1. - is_finished) * p
    return p


def hyp_score(log_probs, length, length_penalty):
    lp = P.pow((5. + P.cast(length, 'float32')) / 6., length_penalty)
    return log_probs / lp


def beam_search_step(state, logits, eos_id, beam_width, is_first_step,
                     length_penalty):
    """logits.shape == [B*W, V]"""
    _, vocab_size = logits.shape

    bsz, beam_width = state.log_probs.shape
    onehot_eos = P.cast(
        F.one_hot(P.ones([1], 'int64') * eos_id, vocab_size), 'int64')  #[1, V]

    probs = P.log(F.softmax(logits))  #[B*W, V]
    probs = mask_prob(probs, onehot_eos, state.finished)  #[B*W, V]
    allprobs = P.reshape(state.log_probs, [-1, 1]) + probs  #[B*W, V]

    not_finished = 1 - P.reshape(state.finished, [-1, 1])  #[B*W,1]
    not_eos = 1 - onehot_eos
    length_to_add = not_finished * not_eos  #[B*W,V]
    alllen = P.reshape(state.lengths, [-1, 1]) + length_to_add

    allprobs = P.reshape(allprobs, [-1, beam_width * vocab_size])
    alllen = P.reshape(alllen, [-1, beam_width * vocab_size])
    allscore = hyp_score(allprobs, alllen, length_penalty)
    if is_first_step:
        allscore = P.reshape(
            allscore,
            [bsz, beam_width, -1])[:, 0, :]  # first step only consiter beam 0
    scores, idx = P.topk(allscore, k=beam_width)  #[B, W]
    next_beam_id = idx // vocab_size  #[B, W]
    next_word_id = idx % vocab_size

    gather_idx = P.concat(
        [P.nonzero(idx != -1)[:, :1], P.reshape(idx, [-1, 1])], 1)
    next_probs = P.reshape(P.gather_nd(allprobs, gather_idx), idx.shape)
    next_len = P.reshape(P.gather_nd(alllen, gather_idx), idx.shape)

    gather_idx = P.concat([
        P.nonzero(next_beam_id != -1)[:, :1], P.reshape(next_beam_id, [-1, 1])
    ], 1)
    next_finished = P.reshape(
        P.gather_nd(state.finished, gather_idx), state.finished.
        shape)  #[gather new beam state according to new beam id]
    #log.debug(gather_idx.numpy())
    #log.debug(state.finished.numpy())
    #log.debug(next_finished.numpy())

    next_finished += P.cast(next_word_id == eos_id, 'int64')
    next_finished = P.cast(next_finished > 0, 'int64')

    #log.debug(next_word_id.numpy())
    #log.debug(next_beam_id.numpy())
    next_state = BeamSearchState(
        log_probs=next_probs, lengths=next_len, finished=next_finished)
    output = BeamSearchOutput(
        scores=scores,
        predicted_ids=next_word_id,
        beam_parent_ids=next_beam_id)

    return output, next_state


def beam_search_infilling(model,
                          q_ids,
                          q_sids,
                          sos_id,
                          eos_id,
                          attn_id,
                          max_encode_len=640,
                          max_decode_len=100,
                          beam_width=5,
                          tgt_type_id=3,
                          length_penalty=1.0):
    model.eval()
    with P.no_grad():
        #log.debug(q_ids.numpy().tolist())
        _, __, info = model(q_ids, q_sids)
        d_batch, d_seqlen = q_ids.shape

        state = BeamSearchState(
            log_probs=P.zeros([d_batch, beam_width], 'float32'),
            lengths=P.zeros([d_batch, beam_width], 'int64'),
            finished=P.zeros([d_batch, beam_width], 'int64'))
        outputs = []

        def reorder_(t, parent_id):
            """reorder cache according to parent beam id"""
            gather_idx = P.nonzero(
                parent_id != -1)[:, 0] * beam_width + P.reshape(parent_id,
                                                                [-1])
            t = P.gather(t, gather_idx)
            return t

        def tile_(t, times):
            _shapes = list(t.shape[1:])
            ret = P.reshape(
                P.tile(
                    P.unsqueeze(t, [1]), [
                        1,
                        times,
                    ] + [1, ] * len(_shapes)), [-1, ] + _shapes)
            return ret

        cached_k, cached_v = info['caches']
        cached_k = [tile_(k, beam_width) for k in cached_k]
        cached_v = [tile_(v, beam_width) for v in cached_v]
        past_cache = (cached_k, cached_v)

        q_ids = tile_(q_ids, beam_width)
        seqlen = P.cast(q_ids != 0, 'int64').sum(1, keepdim=True)
        #log.debug(q_ids.shape)

        cls_ids = P.ones([d_batch * beam_width], dtype='int64') * sos_id
        attn_ids = P.ones(
            [d_batch * beam_width], dtype='int64') * attn_id  # SOS
        ids = P.stack([cls_ids, attn_ids], -1)
        for step in range(max_decode_len):
            #log.debug('decode step %d' % step)
            bias = gen_bias(q_ids, ids, step)
            pos_ids = P.to_tensor(
                np.tile(
                    np.array(
                        [[step, step + 1]], dtype=np.int64),
                    [d_batch * beam_width, 1]))
            pos_ids += seqlen
            _, logits, info = model(
                ids,
                P.ones_like(ids) * tgt_type_id,
                pos_ids=pos_ids,
                attn_bias=bias,
                past_cache=past_cache)

            output, state = beam_search_step(
                state,
                logits[:, 1],
                eos_id=eos_id,
                beam_width=beam_width,
                is_first_step=(step == 0),
                length_penalty=length_penalty)
            outputs.append(output)

            past_cached_k, past_cached_v = past_cache
            cached_k, cached_v = info['caches']
            cached_k = [
                reorder_(
                    P.concat([pk, k[:, :1, :]], 1), output.beam_parent_ids)
                for pk, k in zip(past_cached_k, cached_k)
            ]  # concat cached
            cached_v = [
                reorder_(
                    P.concat([pv, v[:, :1, :]], 1), output.beam_parent_ids)
                for pv, v in zip(past_cached_v, cached_v)
            ]
            past_cache = (cached_k, cached_v)

            pred_ids_flatten = P.reshape(output.predicted_ids,
                                         [d_batch * beam_width])
            ids = P.stack([pred_ids_flatten, attn_ids], 1)

            if state.finished.numpy().all():
                #log.debug('exit because all done')
                break
            #if step == 1: break

        final_ids = P.stack([o.predicted_ids for o in outputs], 0)
        final_parent_ids = P.stack([o.beam_parent_ids for o in outputs], 0)
        final_ids = P.fluid.layers.gather_tree(
            final_ids, final_parent_ids)[:, :, 0]  #pick best beam
        final_ids = P.transpose(
            P.reshape(final_ids, [-1, d_batch * 1]), [1, 0])
    return final_ids


en_patten = re.compile(r'^[a-zA-Z0-9]*$')


def post_process(token):
    if token.startswith('##'):
        ret = token[2:]
    else:
        if en_patten.match(token):
            ret = ' ' + token
        else:
            ret = token
    return ret


if __name__ == '__main__':
    sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-8')
    sys.stderr = io.TextIOWrapper(sys.stderr.buffer, encoding='utf-8')

    parser = argparse.ArgumentParser('seq2seq model with ERNIE')
    parser.add_argument(
        '--from_pretrained',
        type=Path,
        required=True,
        help='pretrained model directory or tag')
    parser.add_argument('--bsz', type=int, default=8, help='batchsize')
    parser.add_argument('--max_encode_len', type=int, default=640)
    parser.add_argument('--max_decode_len', type=int, default=120)
    parser.add_argument('--tgt_type_id', type=int, default=3)
    parser.add_argument('--beam_width', type=int, default=5)
    parser.add_argument(
        '--attn_token',
        type=str,
        default='[ATTN]',
        help='if [ATTN] not in vocab, you can specified [MAKK] as attn-token')
    parser.add_argument('--length_penalty', type=float, default=1.0)
    parser.add_argument(
        '--save_dir', type=str, required=True, help='model dir to be loaded')

    args = parser.parse_args()

    env = P.distributed.ParallelEnv()

    ernie = ErnieModelForGeneration.from_pretrained(
        args.from_pretrained, name='')
    tokenizer = ErnieTokenizer.from_pretrained(
        args.from_pretrained, mask_token=None)
    rev_dict = {v: k for k, v in tokenizer.vocab.items()}
    rev_dict[tokenizer.pad_id] = ''  # replace [PAD]
    rev_dict[tokenizer.unk_id] = ''  # replace [PAD]

    sd = P.load(args.save_dir)
    ernie.set_state_dict(sd)

    def map_fn(src_ids):
        src_ids = src_ids[:args.max_encode_len]
        src_ids, src_sids = tokenizer.build_for_ernie(src_ids)
        return (src_ids, src_sids)

    feature_column = propeller.data.FeatureColumns([
        propeller.data.TextColumn(
            'seg_a',
            unk_id=tokenizer.unk_id,
            vocab_dict=tokenizer.vocab,
            tokenizer=tokenizer.tokenize),
    ])
    dataset = feature_column.build_dataset_from_stdin('predict').map(
        map_fn).padded_batch(args.bsz)

    for step, (encoder_ids, encoder_sids) in enumerate(dataset):
        #result_ids = greedy_search_infilling(ernie, P.to_tensor(encoder_ids), P.to_tensor(encoder_sids),
        #       eos_id=tokenizer.sep_id,
        #       sos_id=tokenizer.cls_id,
        #       attn_id=tokenizer.vocab[args.attn_id],
        #    max_decode_len=args.max_decode_len,
        #    max_encode_len=args.max_encode_len,
        #    beam_width=args.beam_width,
        #    tgt_type_id=args.tgt_type_id)
        result_ids = beam_search_infilling(
            ernie,
            P.to_tensor(encoder_ids),
            P.to_tensor(encoder_sids),
            eos_id=tokenizer.sep_id,
            sos_id=tokenizer.cls_id,
            attn_id=tokenizer.vocab[args.attn_token],
            max_decode_len=args.max_decode_len,
            max_encode_len=args.max_encode_len,
            beam_width=args.beam_width,
            length_penalty=args.length_penalty,
            tgt_type_id=args.tgt_type_id)

        output_str = rev_lookup(result_ids.numpy())
        for ostr in output_str.tolist():
            if '[SEP]' in ostr:
                ostr = ostr[:ostr.index('[SEP]')]

            ostr = ''.join(map(post_process, ostr))
            ostr = ostr.strip()
            print(ostr)