# Copyright (c) Facebook, Inc. and its affiliates. # # This source code is licensed under the MIT license found in the # LICENSE file in the root directory of this source tree. import math import sys from typing import Dict, List, Optional import torch from torch import Tensor from fairseq.sequence_generator import EnsembleModel as EnsembleModelBase from fairseq.sequence_generator import SequenceGenerator as SequenceGeneratorBase class SequenceGenerator(SequenceGeneratorBase): def __init__( self, models, tgt_dict, beam_size=1, max_len_a=0, max_len_b=200, max_len=0, min_len=1, normalize_scores=True, len_penalty=1.0, unk_penalty=0.0, temperature=1.0, match_source_len=False, no_repeat_ngram_size=0, search_strategy=None, eos=None, symbols_to_strip_from_output=None, lm_model=None, lm_weight=1.0, tokens_to_suppress=(), use_incremental_states=False, ): """Generates translations of a given source sentence. Args: models (List[~fairseq.models.FairseqModel]): ensemble of models, currently support fairseq.models.TransformerModel for scripting beam_size (int, optional): beam width (default: 1) max_len_a/b (int, optional): generate sequences of maximum length ax + b, where x is the source length max_len (int, optional): the maximum length of the generated output (not including end-of-sentence) min_len (int, optional): the minimum length of the generated output (not including end-of-sentence) normalize_scores (bool, optional): normalize scores by the length of the output (default: True) len_penalty (float, optional): length penalty, where <1.0 favors shorter, >1.0 favors longer sentences (default: 1.0) unk_penalty (float, optional): unknown word penalty, where <0 produces more unks, >0 produces fewer (default: 0.0) temperature (float, optional): temperature, where values >1.0 produce more uniform samples and values <1.0 produce sharper samples (default: 1.0) match_source_len (bool, optional): outputs should match the source length (default: False) """ super().__init__( models=models, tgt_dict=tgt_dict, beam_size=beam_size, max_len_a=max_len_a, max_len_b=max_len_b, max_len=max_len, min_len=min_len, normalize_scores=normalize_scores, len_penalty=len_penalty, unk_penalty=unk_penalty, temperature=temperature, match_source_len=match_source_len, no_repeat_ngram_size=no_repeat_ngram_size, search_strategy=search_strategy, eos=eos, symbols_to_strip_from_output=symbols_to_strip_from_output, lm_model=lm_model, lm_weight=lm_weight, tokens_to_suppress=tokens_to_suppress, ) if isinstance(models, EnsembleModel): self.model = models else: self.model = EnsembleModel(models) self.model.set_decoder_beam_size(self.beam_size) self.model.eval() self.use_incremental_states = use_incremental_states self.incremental_states = None if not self.use_incremental_states: self.model.has_incremental = False def reset_incremental_states(self): self.incremental_states = None def _generate( self, sample: Dict[str, Dict[str, Tensor]], prefix_tokens: Optional[Tensor] = None, constraints: Optional[Tensor] = None, bos_token: Optional[int] = None, ): net_input = sample["net_input"] if "src_tokens" in net_input: src_tokens = net_input["src_tokens"] # length of the source text being the character length except EndOfSentence and pad # if src_lengths exists in net_input (speech_to_text dataset case), then use it if "src_lengths" in net_input: src_lengths = net_input["src_lengths"] else: src_lengths = ( (src_tokens.ne(self.eos) & src_tokens.ne(self.pad)) .long() .sum(dim=1) ) elif "source" in net_input: src_tokens = net_input["source"] src_lengths = ( net_input["padding_mask"].size(-1) - net_input["padding_mask"].sum(-1) if net_input["padding_mask"] is not None else torch.tensor(src_tokens.size(-1)).to(src_tokens) ) elif "features" in net_input: src_tokens = net_input["features"] src_lengths = ( net_input["padding_mask"].size(-1) - net_input["padding_mask"].sum(-1) if net_input["padding_mask"] is not None else torch.tensor(src_tokens.size(-1)).to(src_tokens) ) else: raise Exception( "expected src_tokens or source in net input. input keys: " + str(net_input.keys()) ) if constraints is not None and not self.search.supports_constraints: raise NotImplementedError( "Target-side constraints were provided, but search method doesn't support them" ) # Initialize constraints, when active self.search.init_constraints(constraints, self.beam_size) # compute the encoder output for each beam with torch.autograd.profiler.record_function("EnsembleModel: forward_encoder"): encoder_outs = self.model.forward_encoder(net_input) finalized = self.generate_decoder( encoder_outs, src_tokens, src_lengths, sample, prefix_tokens, constraints, bos_token, ) return finalized def generate_decoder( self, encoder_outs, src_tokens, src_lengths, sample: Dict[str, Dict[str, Tensor]], prefix_tokens: Optional[Tensor] = None, constraints: Optional[Tensor] = None, bos_token: Optional[int] = None, aux_task_name="", encoder_outs_aug: Optional[ Tensor ] = None, # an additional/augmented encoder_outs max_new_tokens=-1, **kwargs, ): if self.use_incremental_states: if self.incremental_states is None: incremental_states = torch.jit.annotate( List[Dict[str, Dict[str, Optional[Tensor]]]], [ torch.jit.annotate(Dict[str, Dict[str, Optional[Tensor]]], {}) for i in range(self.model.models_size) ], ) self.incremental_states = incremental_states else: incremental_states = self.incremental_states else: incremental_states = None # bsz: total number of sentences in beam # Note that src_tokens may have more than 2 dimensions (i.e. audio features) bsz, src_len = src_tokens.size()[:2] beam_size = self.beam_size decoder_name = f"{aux_task_name}_decoder" if aux_task_name else "decoder" start = prefix_tokens.size(-1) if prefix_tokens is not None else 0 if max_new_tokens == -1: max_len: int = -1 if self.match_source_len: max_len = src_lengths.max().item() else: max_len = min( int(self.max_len_a * src_len + self.max_len_b), self.max_len - 1, ) else: max_len = start + max_new_tokens assert ( self.min_len <= max_len ), "min_len cannot be larger than max_len, please adjust these!" # placeholder of indices for bsz * beam_size to hold tokens and accumulative scores new_order = ( torch.arange(bsz, device=src_tokens.device) .view(-1, 1) .repeat(1, beam_size) .view(-1) ) new_order = new_order.long() encoder_outs = self.model.reorder_encoder_out(encoder_outs, new_order) # ensure encoder_outs is a List. assert encoder_outs is not None if encoder_outs_aug is not None: encoder_outs_aug = self.model.reorder_encoder_out( encoder_outs_aug, new_order ) # initialize buffers scores = torch.zeros( bsz * beam_size, max_len + 1, device=src_tokens.device ).float() # +1 for eos; pad is never chosen for scoring if prefix_tokens is None: tokens = ( torch.zeros(bsz * beam_size, max_len + 2, device=src_tokens.device) .long() .fill_(self.pad) ) # +2 for eos and pad tokens[:, 0] = self.eos if bos_token is None else bos_token else: tokens = torch.cat( ( torch.full( (bsz, 1), self.eos, dtype=torch.long, device=src_tokens.device ), prefix_tokens, ), dim=-1, ) tokens = tokens.repeat(beam_size, 1) if tokens.size(-1) < max_len + 2: tokens = torch.cat( ( tokens, torch.zeros( tokens.size(0), max_len + 2 - tokens.size(-1), device=src_tokens.device, ).long(), ), dim=-1, ) attn: Optional[Tensor] = None # A list that indicates candidates that should be ignored. # For example, suppose we're sampling and have already finalized 2/5 # samples. Then cands_to_ignore would mark 2 positions as being ignored, # so that we only finalize the remaining 3 samples. cands_to_ignore = torch.zeros(bsz, beam_size, device=src_tokens.device).eq( -1 ) # forward and backward-compatible False mask # list of completed sentences finalized = torch.jit.annotate( List[List[Dict[str, Tensor]]], [torch.jit.annotate(List[Dict[str, Tensor]], []) for i in range(bsz)], ) # contains lists of dictionaries of infomation about the hypothesis being finalized at each step # a boolean array indicating if the sentence at the index is finished or not finished = [False for i in range(bsz)] num_remaining_sent = bsz # number of sentences remaining # number of candidate hypos per step cand_size = 2 * beam_size # 2 x beam size in case half are EOS # offset arrays for converting between different indexing schemes bbsz_offsets = ( (torch.arange(0, bsz, device=src_tokens.device) * beam_size) .unsqueeze(1) .type_as(tokens) ) cand_offsets = torch.arange(0, cand_size, device=src_tokens.device).type_as( tokens ) reorder_state: Optional[Tensor] = None batch_idxs: Optional[Tensor] = None original_batch_idxs: Optional[Tensor] = None if "id" in sample and isinstance(sample["id"], Tensor): original_batch_idxs = sample["id"] else: original_batch_idxs = torch.arange(0, bsz).type_as(tokens) for step in range(start, max_len + 1): # one extra step for EOS marker # reorder decoder internal states based on the prev choice of beams if reorder_state is not None: if batch_idxs is not None: # update beam indices to take into account removed sentences corr = batch_idxs - torch.arange(batch_idxs.numel()).type_as( batch_idxs ) reorder_state.view(-1, beam_size).add_( corr.unsqueeze(-1) * beam_size ) original_batch_idxs = original_batch_idxs[batch_idxs] self.model.reorder_incremental_state( incremental_states, reorder_state, decoder_name ) encoder_outs = self.model.reorder_encoder_out( encoder_outs, reorder_state ) if encoder_outs_aug is not None: encoder_outs_aug = self.model.reorder_encoder_out( encoder_outs_aug, reorder_state ) with torch.autograd.profiler.record_function( "EnsembleModel: forward_decoder" ): lprobs, avg_attn_scores = self.model.forward_decoder( tokens[:, : step + 1], encoder_outs, incremental_states, self.temperature, decoder_name=decoder_name, encoder_outs_aug=encoder_outs_aug, **kwargs, ) if self.lm_model is not None and not aux_task_name: lm_out = self.lm_model(tokens[:, : step + 1]) probs = self.lm_model.get_normalized_probs( lm_out, log_probs=True, sample=None ) probs = probs[:, -1, :] * self.lm_weight lprobs += probs lprobs[lprobs != lprobs] = torch.tensor(-math.inf, device=lprobs.device) lprobs[:, self.pad] = -math.inf # never select pad lprobs[:, self.unk] -= self.unk_penalty # apply unk penalty # handle max length constraint if step >= max_len: lprobs[:, : self.eos] = -math.inf lprobs[:, self.eos + 1 :] = -math.inf # handle prefix tokens (possibly with different lengths) if ( prefix_tokens is not None and step < prefix_tokens.size(1) and step < max_len ): lprobs, tokens, scores = self._prefix_tokens( step, lprobs, scores, tokens, prefix_tokens, beam_size ) else: if step < self.min_len: # minimum length constraint (does not apply if using prefix_tokens) lprobs[:, self.eos] = -math.inf if self.token_indices_to_suppress is not None: lprobs[:, self.token_indices_to_suppress] = -math.inf # Record attention scores, only support avg_attn_scores is a Tensor if avg_attn_scores is not None: if attn is None: attn = torch.empty( bsz * beam_size, avg_attn_scores.size(1), max_len + 2, device=scores.device, ) attn[:, :, step + 1].copy_(avg_attn_scores) scores = scores.type_as(lprobs) eos_bbsz_idx = torch.empty(0, device=tokens.device).type_as( tokens ) # indices of hypothesis ending with eos (finished sentences) eos_scores = torch.empty(0, device=scores.device).type_as( scores ) # scores of hypothesis ending with eos (finished sentences) if self.should_set_src_lengths: self.search.set_src_lengths(src_lengths) if self.repeat_ngram_blocker is not None: lprobs = self.repeat_ngram_blocker(tokens, lprobs, bsz, beam_size, step) # Shape: (batch, cand_size) cand_scores, cand_indices, cand_beams = self.search.step( step, lprobs.view(bsz, -1, self.vocab_size), scores.view(bsz, beam_size, -1)[:, :, :step], tokens[:, : step + 1], original_batch_idxs, ) # cand_bbsz_idx contains beam indices for the top candidate # hypotheses, with a range of values: [0, bsz*beam_size), # and dimensions: [bsz, cand_size] cand_bbsz_idx = cand_beams.add(bbsz_offsets) # finalize hypotheses that end in eos # Shape of eos_mask: (batch size, beam size) eos_mask = cand_indices.eq(self.eos) & cand_scores.ne(-math.inf) eos_mask[:, :beam_size][cands_to_ignore] = torch.tensor( 0, device=eos_mask.device ).type_as(eos_mask) # only consider eos when it's among the top beam_size indices # Now we know what beam item(s) to finish # Shape: 1d list of absolute-numbered eos_bbsz_idx = torch.masked_select( cand_bbsz_idx[:, :beam_size], mask=eos_mask[:, :beam_size] ) finalized_sents: List[int] = [] if eos_bbsz_idx.numel() > 0: eos_scores = torch.masked_select( cand_scores[:, :beam_size], mask=eos_mask[:, :beam_size] ) finalized_sents = self.finalize_hypos( step, eos_bbsz_idx, eos_scores, tokens, scores, finalized, finished, beam_size, attn, src_lengths, max_len, ) num_remaining_sent -= len(finalized_sents) assert num_remaining_sent >= 0 if num_remaining_sent == 0: break if self.search.stop_on_max_len and step >= max_len: break assert step < max_len, f"{step} < {max_len}" # Remove finalized sentences (ones for which {beam_size} # finished hypotheses have been generated) from the batch. if len(finalized_sents) > 0: new_bsz = bsz - len(finalized_sents) # construct batch_idxs which holds indices of batches to keep for the next pass batch_mask = torch.ones( bsz, dtype=torch.bool, device=cand_indices.device ) batch_mask[finalized_sents] = False # TODO replace `nonzero(as_tuple=False)` after TorchScript supports it batch_idxs = ( torch.arange(bsz, device=cand_indices.device) .type_as(cand_indices) .masked_select(batch_mask) ) # Choose the subset of the hypothesized constraints that will continue self.search.prune_sentences(batch_idxs) eos_mask = eos_mask[batch_idxs] cand_beams = cand_beams[batch_idxs] bbsz_offsets.resize_(new_bsz, 1) cand_bbsz_idx = cand_beams.add(bbsz_offsets) cand_scores = cand_scores[batch_idxs] cand_indices = cand_indices[batch_idxs] if prefix_tokens is not None: prefix_tokens = prefix_tokens[batch_idxs] src_lengths = src_lengths[batch_idxs] cands_to_ignore = cands_to_ignore[batch_idxs] scores = scores.view(bsz, -1)[batch_idxs].view(new_bsz * beam_size, -1) tokens = tokens.view(bsz, -1)[batch_idxs].view(new_bsz * beam_size, -1) if attn is not None: attn = attn.view(bsz, -1)[batch_idxs].view( new_bsz * beam_size, attn.size(1), -1 ) bsz = new_bsz else: batch_idxs = None # Set active_mask so that values > cand_size indicate eos hypos # and values < cand_size indicate candidate active hypos. # After, the min values per row are the top candidate active hypos # Rewrite the operator since the element wise or is not supported in torchscript. eos_mask[:, :beam_size] = ~((~cands_to_ignore) & (~eos_mask[:, :beam_size])) active_mask = torch.add( eos_mask.type_as(cand_offsets) * cand_size, cand_offsets[: eos_mask.size(1)], ) # get the top beam_size active hypotheses, which are just # the hypos with the smallest values in active_mask. # {active_hypos} indicates which {beam_size} hypotheses # from the list of {2 * beam_size} candidates were # selected. Shapes: (batch size, beam size) new_cands_to_ignore, active_hypos = torch.topk( active_mask, k=beam_size, dim=1, largest=False ) # update cands_to_ignore to ignore any finalized hypos. cands_to_ignore = new_cands_to_ignore.ge(cand_size)[:, :beam_size] # Make sure there is at least one active item for each sentence in the batch. assert (~cands_to_ignore).any(dim=1).all() # update cands_to_ignore to ignore any finalized hypos # {active_bbsz_idx} denotes which beam number is continued for each new hypothesis (a beam # can be selected more than once). active_bbsz_idx = torch.gather(cand_bbsz_idx, dim=1, index=active_hypos) active_scores = torch.gather(cand_scores, dim=1, index=active_hypos) active_bbsz_idx = active_bbsz_idx.view(-1) active_scores = active_scores.view(-1) # copy tokens and scores for active hypotheses # Set the tokens for each beam (can select the same row more than once) tokens[:, : step + 1] = torch.index_select( tokens[:, : step + 1], dim=0, index=active_bbsz_idx ) # Select the next token for each of them tokens.view(bsz, beam_size, -1)[:, :, step + 1] = torch.gather( cand_indices, dim=1, index=active_hypos ) if step > 0: scores[:, :step] = torch.index_select( scores[:, :step], dim=0, index=active_bbsz_idx ) scores.view(bsz, beam_size, -1)[:, :, step] = torch.gather( cand_scores, dim=1, index=active_hypos ) # Update constraints based on which candidates were selected for the next beam self.search.update_constraints(active_hypos) # copy attention for active hypotheses if attn is not None: attn[:, :, : step + 2] = torch.index_select( attn[:, :, : step + 2], dim=0, index=active_bbsz_idx ) # reorder incremental state in decoder reorder_state = active_bbsz_idx # sort by score descending for sent in range(len(finalized)): scores = torch.tensor( [float(elem["score"].item()) for elem in finalized[sent]] ) _, sorted_scores_indices = torch.sort(scores, descending=True) finalized[sent] = [finalized[sent][ssi] for ssi in sorted_scores_indices] finalized[sent] = torch.jit.annotate( List[Dict[str, Tensor]], finalized[sent] ) return finalized class EnsembleModel(EnsembleModelBase): """A wrapper around an ensemble of models.""" def __init__(self, models): super().__init__(models) @torch.jit.export def forward_decoder( self, tokens, encoder_outs: List[Dict[str, List[Tensor]]], incremental_states: List[Dict[str, Dict[str, Optional[Tensor]]]], temperature: float = 1.0, decoder_name="decoder", encoder_outs_aug: List[Dict[str, List[Tensor]]] = None, **kwargs, ): log_probs = [] avg_attn: Optional[Tensor] = None encoder_out: Optional[Dict[str, List[Tensor]]] = None encoder_out_aug: Optional[Dict[str, List[Tensor]]] = None for i, model in enumerate(self.models): if self.has_encoder(): encoder_out = encoder_outs[i] if encoder_outs_aug is not None: encoder_out_aug = encoder_outs_aug[i] # decode each model if self.has_incremental_states() and incremental_states is not None: if encoder_out_aug is not None: decoder_out = getattr(model, decoder_name).forward( tokens, encoder_out=encoder_out, encoder_out_aug=encoder_out_aug, incremental_state=incremental_states[i], ) else: decoder_out = getattr(model, decoder_name).forward( tokens, encoder_out=encoder_out, incremental_state=incremental_states[i], **kwargs, ) else: if hasattr(model, decoder_name): decoder_out = getattr(model, decoder_name).forward( tokens, encoder_out=encoder_out ) else: decoder_out = model.forward(tokens) attn: Optional[Tensor] = None decoder_len = len(decoder_out) if decoder_len > 1 and decoder_out[1] is not None: if isinstance(decoder_out[1], Tensor): attn = decoder_out[1] else: attn_holder = decoder_out[1]["attn"] if isinstance(attn_holder, Tensor): attn = attn_holder elif attn_holder is not None: attn = attn_holder[0] if attn is not None: attn = attn[:, -1, :] decoder_out_tuple = ( decoder_out[0][:, -1:, :].div_(temperature), None if decoder_len <= 1 else decoder_out[1], ) probs = getattr(model, decoder_name).get_normalized_probs( decoder_out_tuple, log_probs=True, sample=None ) probs = probs[:, -1, :] if self.models_size == 1: return probs, attn log_probs.append(probs) if attn is not None: if avg_attn is None: avg_attn = attn else: avg_attn.add_(attn) avg_probs = torch.logsumexp(torch.stack(log_probs, dim=0), dim=0) - math.log( self.models_size ) if avg_attn is not None: avg_attn.div_(self.models_size) return avg_probs, avg_attn @torch.jit.export def reorder_incremental_state( self, incremental_states: List[Dict[str, Dict[str, Optional[Tensor]]]], new_order, decoder_name="decoder", ): if not self.has_incremental_states(): return for i, model in enumerate(self.models): getattr(model, decoder_name).reorder_incremental_state_scripting( incremental_states[i], new_order )