add: update generation code
Browse files- generation_utils.py +82 -0
- modeling_baichuan.py +9 -42
- requirements.txt +0 -1
generation_utils.py
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from typing import List
|
2 |
+
from queue import Queue
|
3 |
+
|
4 |
+
import torch
|
5 |
+
|
6 |
+
|
7 |
+
def build_chat_input(model, tokenizer, messages: List[dict], max_new_tokens: int=0):
|
8 |
+
def _parse_messages(messages, split_role="user"):
|
9 |
+
system, rounds = "", []
|
10 |
+
round = []
|
11 |
+
for i, message in enumerate(messages):
|
12 |
+
if message["role"] == "system":
|
13 |
+
assert i == 0
|
14 |
+
system = message["content"]
|
15 |
+
continue
|
16 |
+
if message["role"] == split_role and round:
|
17 |
+
rounds.append(round)
|
18 |
+
round = []
|
19 |
+
round.append(message)
|
20 |
+
if round:
|
21 |
+
rounds.append(round)
|
22 |
+
return system, rounds
|
23 |
+
|
24 |
+
max_new_tokens = max_new_tokens or model.generation_config.max_new_tokens
|
25 |
+
max_input_tokens = model.config.model_max_length - max_new_tokens
|
26 |
+
system, rounds = _parse_messages(messages, split_role="user")
|
27 |
+
system_tokens = tokenizer.encode(system)
|
28 |
+
max_history_tokens = max_input_tokens - len(system_tokens)
|
29 |
+
|
30 |
+
history_tokens = []
|
31 |
+
for round in rounds[::-1]:
|
32 |
+
round_tokens = []
|
33 |
+
for message in round:
|
34 |
+
if message["role"] == "user":
|
35 |
+
round_tokens.append(model.generation_config.user_token_id)
|
36 |
+
else:
|
37 |
+
round_tokens.append(model.generation_config.assistant_token_id)
|
38 |
+
round_tokens.extend(tokenizer.encode(message["content"]))
|
39 |
+
if len(history_tokens) == 0 or len(history_tokens) + len(round_tokens) <= max_history_tokens:
|
40 |
+
history_tokens = round_tokens + history_tokens # concat left
|
41 |
+
if len(history_tokens) < max_history_tokens:
|
42 |
+
continue
|
43 |
+
break
|
44 |
+
|
45 |
+
input_tokens = system_tokens + history_tokens
|
46 |
+
if messages[-1]["role"] != "assistant":
|
47 |
+
input_tokens.append(model.generation_config.assistant_token_id)
|
48 |
+
input_tokens = input_tokens[-max_input_tokens:] # truncate left
|
49 |
+
return torch.LongTensor([input_tokens]).to(model.device)
|
50 |
+
|
51 |
+
|
52 |
+
class TextIterStreamer:
|
53 |
+
def __init__(self, tokenizer, skip_prompt=False, skip_special_tokens=False):
|
54 |
+
self.tokenizer = tokenizer
|
55 |
+
self.skip_prompt = skip_prompt
|
56 |
+
self.skip_special_tokens = skip_special_tokens
|
57 |
+
self.tokens = []
|
58 |
+
self.text_queue = Queue()
|
59 |
+
self.next_tokens_are_prompt = True
|
60 |
+
|
61 |
+
def put(self, value):
|
62 |
+
if self.skip_prompt and self.next_tokens_are_prompt:
|
63 |
+
self.next_tokens_are_prompt = False
|
64 |
+
else:
|
65 |
+
if len(value.shape) > 1:
|
66 |
+
value = value[0]
|
67 |
+
self.tokens.extend(value.tolist())
|
68 |
+
self.text_queue.put(
|
69 |
+
self.tokenizer.decode(self.tokens, skip_special_tokens=self.skip_special_tokens))
|
70 |
+
|
71 |
+
def end(self):
|
72 |
+
self.text_queue.put(None)
|
73 |
+
|
74 |
+
def __iter__(self):
|
75 |
+
return self
|
76 |
+
|
77 |
+
def __next__(self):
|
78 |
+
value = self.text_queue.get()
|
79 |
+
if value is None:
|
80 |
+
raise StopIteration()
|
81 |
+
else:
|
82 |
+
return value
|
modeling_baichuan.py
CHANGED
@@ -1,6 +1,7 @@
|
|
1 |
# Copyright (c) 2023, Baichuan Intelligent Technology. All rights reserved.
|
2 |
|
3 |
import math
|
|
|
4 |
from typing import List, Optional, Tuple, Union
|
5 |
|
6 |
import torch
|
@@ -13,6 +14,7 @@ from transformers.utils import logging
|
|
13 |
from transformers.generation.utils import GenerationConfig
|
14 |
|
15 |
from .configuration_baichuan import BaichuanConfig
|
|
|
16 |
|
17 |
logger = logging.get_logger(__name__)
|
18 |
|
@@ -552,54 +554,19 @@ class BaichuanForCausalLM(BaichuanPreTrainedModel):
|
|
552 |
)
|
553 |
return self
|
554 |
|
555 |
-
def _build_chat_input(self, tokenizer, messages: List[dict], max_new_tokens: int=0):
|
556 |
-
max_new_tokens = max_new_tokens or self.generation_config.max_new_tokens
|
557 |
-
max_input_tokens = self.config.model_max_length - max_new_tokens
|
558 |
-
max_input_tokens = max(self.config.model_max_length // 2, max_input_tokens)
|
559 |
-
total_input, round_input = [], []
|
560 |
-
for i, message in enumerate(messages[::-1]):
|
561 |
-
content_tokens = tokenizer.encode(message['content'])
|
562 |
-
if message['role'] == 'user':
|
563 |
-
round_input = [self.generation_config.user_token_id] + content_tokens + round_input
|
564 |
-
if total_input and len(total_input) + len(round_input) > max_input_tokens:
|
565 |
-
break
|
566 |
-
else:
|
567 |
-
total_input = round_input + total_input
|
568 |
-
if len(total_input) >= max_input_tokens:
|
569 |
-
break
|
570 |
-
else:
|
571 |
-
round_input = []
|
572 |
-
elif message['role'] == 'assistant':
|
573 |
-
round_input = [
|
574 |
-
self.generation_config.assistant_token_id
|
575 |
-
] + content_tokens + round_input
|
576 |
-
else:
|
577 |
-
raise ValueError(f"message role not supported yet: {message['role']}")
|
578 |
-
total_input = total_input[-max_input_tokens:] # truncate left
|
579 |
-
total_input.append(self.generation_config.assistant_token_id)
|
580 |
-
total_input = torch.LongTensor([total_input]).to(self.device)
|
581 |
-
return total_input
|
582 |
-
|
583 |
@torch.no_grad()
|
584 |
def chat(self, tokenizer, messages: List[dict], stream=False,
|
585 |
generation_config: Optional[GenerationConfig]=None):
|
586 |
generation_config = generation_config or self.generation_config
|
587 |
-
input_ids = self
|
588 |
if stream:
|
589 |
-
|
590 |
-
self.
|
591 |
-
|
592 |
-
|
593 |
-
|
594 |
-
|
595 |
-
outputs = []
|
596 |
-
for token in self.generate(input_ids, generation_config=stream_config):
|
597 |
-
outputs.append(token.item())
|
598 |
-
yield tokenizer.decode(outputs, skip_special_tokens=True)
|
599 |
-
|
600 |
-
return stream_generator()
|
601 |
else:
|
602 |
-
self.__class__.generate = PreTrainedModel.generate # disable stream
|
603 |
outputs = self.generate(input_ids, generation_config=generation_config)
|
604 |
response = tokenizer.decode(outputs[0][len(input_ids[0]):], skip_special_tokens=True)
|
605 |
return response
|
|
|
1 |
# Copyright (c) 2023, Baichuan Intelligent Technology. All rights reserved.
|
2 |
|
3 |
import math
|
4 |
+
from threading import Thread
|
5 |
from typing import List, Optional, Tuple, Union
|
6 |
|
7 |
import torch
|
|
|
14 |
from transformers.generation.utils import GenerationConfig
|
15 |
|
16 |
from .configuration_baichuan import BaichuanConfig
|
17 |
+
from .generation_utils import build_chat_input, TextIterStreamer
|
18 |
|
19 |
logger = logging.get_logger(__name__)
|
20 |
|
|
|
554 |
)
|
555 |
return self
|
556 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
557 |
@torch.no_grad()
|
558 |
def chat(self, tokenizer, messages: List[dict], stream=False,
|
559 |
generation_config: Optional[GenerationConfig]=None):
|
560 |
generation_config = generation_config or self.generation_config
|
561 |
+
input_ids = build_chat_input(self, tokenizer, messages, generation_config.max_new_tokens)
|
562 |
if stream:
|
563 |
+
streamer = TextIterStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
|
564 |
+
Thread(target=self.generate, kwargs=dict(
|
565 |
+
inputs=input_ids, streamer=streamer,
|
566 |
+
generation_config=generation_config,
|
567 |
+
)).start()
|
568 |
+
return streamer
|
|
|
|
|
|
|
|
|
|
|
|
|
569 |
else:
|
|
|
570 |
outputs = self.generate(input_ids, generation_config=generation_config)
|
571 |
response = tokenizer.decode(outputs[0][len(input_ids[0]):], skip_special_tokens=True)
|
572 |
return response
|
requirements.txt
CHANGED
@@ -3,4 +3,3 @@ colorama
|
|
3 |
cpm_kernels
|
4 |
sentencepiece
|
5 |
streamlit
|
6 |
-
transformers_stream_generator
|
|
|
3 |
cpm_kernels
|
4 |
sentencepiece
|
5 |
streamlit
|
|