Witllm/qwen/qwen_generation_utils.py

295 lines
11 KiB
Python
Raw Normal View History

2024-01-03 21:03:27 +08:00
# Copyright (c) Alibaba Cloud.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
"""Generation support."""
from typing import Tuple, List, Union, Iterable
import numpy as np
import torch
import torch.nn.functional as F
from transformers import PreTrainedTokenizer
from transformers import logging
from transformers.generation import LogitsProcessor
logger = logging.get_logger(__name__)
# Types.
HistoryType = List[Tuple[str, str]]
TokensType = List[int]
BatchTokensType = List[List[int]]
def pad_batch(batch: BatchTokensType, pad_id: int, seq_length: int) -> BatchTokensType:
for tokens in batch:
context_length = len(tokens)
if context_length < seq_length:
tokens.extend([pad_id] * (seq_length - context_length))
return batch
def get_ltor_masks_and_position_ids(
data,
eod_token,
reset_position_ids,
reset_attention_mask,
eod_mask_loss,
):
"""Build masks and position id for left to right model."""
# Extract batch size and sequence length.
micro_batch_size, seq_length = data.size()
# Attention mask (lower triangular).
if reset_attention_mask:
att_mask_batch = micro_batch_size
else:
att_mask_batch = 1
2024-01-22 20:57:27 +08:00
attention_mask = torch.tril(torch.ones((att_mask_batch, seq_length, seq_length), device=data.device)).view(
att_mask_batch, 1, seq_length, seq_length
)
2024-01-03 21:03:27 +08:00
# Loss mask.
loss_mask = torch.ones(data.size(), dtype=torch.float, device=data.device)
if eod_mask_loss:
loss_mask[data == eod_token] = 0.0
# Position ids.
position_ids = torch.arange(seq_length, dtype=torch.long, device=data.device)
position_ids = position_ids.unsqueeze(0).expand_as(data)
# We need to clone as the ids will be modifed based on batch index.
if reset_position_ids:
position_ids = position_ids.clone()
if reset_position_ids or reset_attention_mask:
# Loop through the batches:
for b in range(micro_batch_size):
# Find indecies where EOD token is.
eod_index = position_ids[b, data[b] == eod_token]
# Detach indecies from positions if going to modify positions.
if reset_position_ids:
eod_index = eod_index.clone()
# Loop through EOD indecies:
prev_index = 0
for j in range(eod_index.size()[0]):
i = eod_index[j]
# Mask attention loss.
if reset_attention_mask:
attention_mask[b, 0, (i + 1) :, : (i + 1)] = 0
# Reset positions.
if reset_position_ids:
position_ids[b, (i + 1) :] -= i + 1 - prev_index
prev_index = i + 1
# Convert attention mask to binary:
attention_mask = attention_mask < 0.5
return attention_mask, loss_mask, position_ids
def get_batch(context_tokens: torch.LongTensor, eod_id: int):
"""Generate batch from context tokens."""
# Move to GPU.
tokens = context_tokens.contiguous().to(context_tokens.device)
# Get the attention mask and postition ids.
attention_mask, _, position_ids = get_ltor_masks_and_position_ids(
tokens,
eod_id,
reset_position_ids=False,
reset_attention_mask=False,
eod_mask_loss=False,
)
return tokens, attention_mask, position_ids
2024-01-22 20:57:27 +08:00
2024-01-03 21:03:27 +08:00
def make_context(
tokenizer: PreTrainedTokenizer,
query: str,
2024-01-10 19:35:46 +08:00
query_assistant: str = "",
2024-01-03 21:03:27 +08:00
history: List[Tuple[str, str]] = None,
system: str = "",
2024-01-22 20:57:27 +08:00
max_window_size: int = 6144,
2024-01-03 21:03:27 +08:00
):
if history is None:
history = []
2024-01-10 19:35:46 +08:00
im_start, im_end = "<|im_start|>", "<|im_end|>"
im_start_tokens = [tokenizer.im_start_id]
im_end_tokens = [tokenizer.im_end_id]
nl_tokens = tokenizer.encode("\n")
def _tokenize_str(role, content):
2024-01-22 20:57:27 +08:00
return f"{role}\n{content}", tokenizer.encode(role, allowed_special=set()) + nl_tokens + tokenizer.encode(
content, allowed_special=set()
)
2024-01-10 19:35:46 +08:00
system_text, system_tokens_part = _tokenize_str("system", system)
system_tokens = im_start_tokens + system_tokens_part + im_end_tokens
2024-01-22 20:57:27 +08:00
assistant_tokens = tokenizer.encode(query_assistant, allowed_special=set())
2024-01-10 19:35:46 +08:00
raw_text = ""
context_tokens = []
for turn_query, turn_response in reversed(history):
query_text, query_tokens_part = _tokenize_str("user", turn_query)
query_tokens = im_start_tokens + query_tokens_part + im_end_tokens
2024-01-22 20:57:27 +08:00
response_text, response_tokens_part = _tokenize_str("assistant", turn_response)
2024-01-10 19:35:46 +08:00
response_tokens = im_start_tokens + response_tokens_part + im_end_tokens
2024-01-03 21:03:27 +08:00
2024-01-10 19:35:46 +08:00
next_context_tokens = nl_tokens + query_tokens + nl_tokens + response_tokens
2024-01-22 20:57:27 +08:00
prev_chat = f"\n{im_start}{query_text}{im_end}\n{im_start}{response_text}{im_end}"
2024-01-03 21:03:27 +08:00
2024-01-22 20:57:27 +08:00
current_context_size = len(system_tokens) + len(next_context_tokens) + len(context_tokens)
2024-01-10 19:35:46 +08:00
if current_context_size < max_window_size:
context_tokens = next_context_tokens + context_tokens
raw_text = prev_chat + raw_text
else:
2024-01-03 21:03:27 +08:00
break
2024-01-10 19:35:46 +08:00
context_tokens = system_tokens + context_tokens
raw_text = f"{im_start}{system_text}{im_end}" + raw_text
context_tokens += (
nl_tokens
+ im_start_tokens
+ _tokenize_str("user", query)[1]
+ im_end_tokens
+ nl_tokens
+ im_start_tokens
+ tokenizer.encode("assistant")
+ nl_tokens
+ assistant_tokens
)
raw_text += f"\n{im_start}user\n{query}{im_end}\n{im_start}assistant\n{query_assistant}"
2024-01-03 21:03:27 +08:00
2024-01-10 19:35:46 +08:00
return raw_text, context_tokens
2024-01-03 21:03:27 +08:00
2024-01-22 20:57:27 +08:00
2024-01-03 21:03:27 +08:00
def decode_tokens(
tokens: Union[torch.LongTensor, TokensType],
tokenizer: PreTrainedTokenizer,
2024-01-22 20:57:27 +08:00
raw_text_len: int = 0,
context_length: int = 0,
errors: str = "replace",
2024-01-03 21:03:27 +08:00
) -> str:
if torch.is_tensor(tokens):
tokens = tokens.cpu().numpy().tolist()
2024-01-10 19:35:46 +08:00
end_reason = f"Gen length {len(tokens)}"
eod_token_idx = context_length
for eod_token_idx in range(context_length, len(tokens)):
if tokens[eod_token_idx] in [tokenizer.im_start_id, tokenizer.im_end_id]:
end_reason = f"Gen {tokenizer.decode([tokens[eod_token_idx]])!r}"
break
decoded = tokenizer.decode(tokens, errors=errors)
decode_tokens = tokenizer.decode(tokens[:eod_token_idx], errors=errors)
trim_decode_tokens = decode_tokens[raw_text_len:]
trim_decode_tokens = trim_decode_tokens.strip()
return decoded, trim_decode_tokens, end_reason
2024-01-03 21:03:27 +08:00
class StopWordsLogitsProcessor(LogitsProcessor):
"""
:class:`transformers.LogitsProcessor` that enforces that when specified sequences appear, stop geration.
Args:
stop_words_ids (:obj:`List[List[int]]`):
List of list of token ids of stop ids. In order to get the tokens of the words
that should not appear in the generated text, use :obj:`tokenizer(bad_word,
add_prefix_space=True).input_ids`.
eos_token_id (:obj:`int`):
The id of the `end-of-sequence` token.
"""
def __init__(self, stop_words_ids: Iterable[Iterable[int]], eos_token_id: int):
if not isinstance(stop_words_ids, List) or len(stop_words_ids) == 0:
2024-01-22 20:57:27 +08:00
raise ValueError(f"`stop_words_ids` has to be a non-emtpy list, but is {stop_words_ids}.")
2024-01-03 21:03:27 +08:00
if any(not isinstance(bad_word_ids, list) for bad_word_ids in stop_words_ids):
2024-01-22 20:57:27 +08:00
raise ValueError(f"`stop_words_ids` has to be a list of lists, but is {stop_words_ids}.")
2024-01-03 21:03:27 +08:00
if any(
2024-01-22 20:57:27 +08:00
any((not isinstance(token_id, (int, np.integer)) or token_id < 0) for token_id in stop_word_ids)
2024-01-03 21:03:27 +08:00
for stop_word_ids in stop_words_ids
):
raise ValueError(
f"Each list in `stop_words_ids` has to be a list of positive integers, but is {stop_words_ids}."
)
2024-01-22 20:57:27 +08:00
self.stop_words_ids = list(filter(lambda bad_token_seq: bad_token_seq != [eos_token_id], stop_words_ids))
2024-01-03 21:03:27 +08:00
self.eos_token_id = eos_token_id
for stop_token_seq in self.stop_words_ids:
2024-01-22 20:57:27 +08:00
assert len(stop_token_seq) > 0, "Stop words token sequences {} cannot have an empty list".format(
2024-01-03 21:03:27 +08:00
stop_words_ids
)
2024-01-22 20:57:27 +08:00
def __call__(self, input_ids: torch.LongTensor, scores: torch.FloatTensor) -> torch.FloatTensor:
2024-01-03 21:03:27 +08:00
stopped_samples = self._calc_stopped_samples(input_ids)
for i, should_stop in enumerate(stopped_samples):
if should_stop:
scores[i, self.eos_token_id] = float(2**15)
return scores
def _tokens_match(self, prev_tokens: torch.LongTensor, tokens: List[int]) -> bool:
if len(tokens) == 0:
# if bad word tokens is just one token always ban it
return True
elif len(tokens) > len(prev_tokens):
# if bad word tokens are longer then prev input_ids they can't be equal
return False
elif prev_tokens[-len(tokens) :].tolist() == tokens:
# if tokens match
return True
else:
return False
def _calc_stopped_samples(self, prev_input_ids: Iterable[int]) -> Iterable[int]:
stopped_samples = []
for prev_input_ids_slice in prev_input_ids:
match = False
for stop_token_seq in self.stop_words_ids:
if self._tokens_match(prev_input_ids_slice, stop_token_seq):
# if tokens do not match continue
match = True
break
stopped_samples.append(match)
return stopped_samples
def top_k_logits(logits, top_k=0, top_p=0.0, filter_value=-float("Inf")):
"""This function has been mostly taken from huggingface conversational
ai code at
https://medium.com/huggingface/how-to-build-a-state-of-the-art-
conversational-ai-with-transfer-learning-2d818ac26313"""
if top_k > 0:
# Remove all tokens with a probability less than the
# last token of the top-k
indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
logits[indices_to_remove] = filter_value
if top_p > 0.0:
# Cconvert to 1D
sorted_logits, sorted_indices = torch.sort(logits, descending=True, dim=-1)
cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
# Remove tokens with cumulative probability above the threshold
sorted_indices_to_remove = cumulative_probs > top_p
# Shift the indices to the right to keep also the first token
# above the threshold
sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
sorted_indices_to_remove[..., 0] = 0
for i in range(sorted_indices.size(0)):
indices_to_remove = sorted_indices[i][sorted_indices_to_remove[i]]
logits[i][indices_to_remove] = filter_value
return logits
def switch(val1, val2, boolean):
boolean = boolean.type_as(val1)
return (1 - boolean) * val1 + boolean * val2