Spaces:
Running
Running
Husnain
commited on
Commit
•
99c68b6
1
Parent(s):
fcdcbfd
💥 [Fix] Revert chat templates with mannual version as nous-mixtral-8x…
Browse files- messagers/message_composer.py +26 -129
messagers/message_composer.py
CHANGED
@@ -1,23 +1,19 @@
|
|
1 |
import re
|
2 |
from pprint import pprint
|
|
|
|
|
|
|
|
|
3 |
from utils.logger import logger
|
4 |
|
5 |
|
6 |
class MessageComposer:
|
7 |
-
# LINK - apis/chat_api.py#available-models
|
8 |
-
AVALAIBLE_MODELS = [
|
9 |
-
"mixtral-8x7b",
|
10 |
-
"mistral-7b",
|
11 |
-
"openchat-3.5",
|
12 |
-
"nous-mixtral-8x7b",
|
13 |
-
"gemma-7b",
|
14 |
-
]
|
15 |
-
|
16 |
def __init__(self, model: str = None):
|
17 |
-
if model in
|
18 |
self.model = model
|
19 |
else:
|
20 |
self.model = "mixtral-8x7b"
|
|
|
21 |
self.system_roles = ["system"]
|
22 |
self.inst_roles = ["user", "system", "inst"]
|
23 |
self.answer_roles = ["assistant", "bot", "answer", "model"]
|
@@ -51,12 +47,16 @@ class MessageComposer:
|
|
51 |
return concat_messages
|
52 |
|
53 |
def merge(self, messages) -> str:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
# Mistral and Mixtral:
|
55 |
# <s> [INST] Instruction [/INST] Model answer </s> [INST] Follow-up instruction [/INST]
|
56 |
|
57 |
-
# OpenChat:
|
58 |
-
# GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi<|end_of_turn|>GPT4 Correct User: How are you today?<|end_of_turn|>GPT4 Correct Assistant:
|
59 |
-
|
60 |
# Nous Mixtral:
|
61 |
# <|im_start|>system
|
62 |
# You are "Hermes 2".<|im_end|>
|
@@ -64,6 +64,9 @@ class MessageComposer:
|
|
64 |
# Hello, who are you?<|im_end|>
|
65 |
# <|im_start|>assistant
|
66 |
|
|
|
|
|
|
|
67 |
# Google Gemma-it
|
68 |
# <start_of_turn>user
|
69 |
# How does the brain work?<end_of_turn>
|
@@ -145,127 +148,23 @@ class MessageComposer:
|
|
145 |
)
|
146 |
self.merged_str_list.append(f"{self.start_of_turn}model\n")
|
147 |
self.merged_str = "\n".join(self.merged_str_list)
|
|
|
|
|
|
|
148 |
else:
|
149 |
-
|
150 |
-
|
151 |
-
|
152 |
-
for message in self.messages
|
153 |
-
]
|
154 |
)
|
155 |
|
156 |
return self.merged_str
|
157 |
|
158 |
-
def convert_pair_matches_to_messages(self, pair_matches_list):
|
159 |
-
messages = []
|
160 |
-
if len(pair_matches_list) <= 0:
|
161 |
-
messages = [
|
162 |
-
{
|
163 |
-
"role": "user",
|
164 |
-
"content": self.merged_str,
|
165 |
-
}
|
166 |
-
]
|
167 |
-
else:
|
168 |
-
for match in pair_matches_list:
|
169 |
-
inst = match.group("inst")
|
170 |
-
answer = match.group("answer")
|
171 |
-
messages.extend(
|
172 |
-
[
|
173 |
-
{"role": "user", "content": inst.strip()},
|
174 |
-
{"role": "assistant", "content": answer.strip()},
|
175 |
-
]
|
176 |
-
)
|
177 |
-
return messages
|
178 |
-
|
179 |
-
def append_last_instruction_to_messages(self, inst_matches_list, pair_matches_list):
|
180 |
-
if len(inst_matches_list) > len(pair_matches_list):
|
181 |
-
self.messages.extend(
|
182 |
-
[
|
183 |
-
{
|
184 |
-
"role": "user",
|
185 |
-
"content": inst_matches_list[-1].group("inst").strip(),
|
186 |
-
}
|
187 |
-
]
|
188 |
-
)
|
189 |
-
|
190 |
-
def split(self, merged_str) -> list:
|
191 |
-
self.merged_str = merged_str
|
192 |
-
self.messages = []
|
193 |
-
|
194 |
-
if self.model in ["mixtral-8x7b", "mistral-7b"]:
|
195 |
-
pair_pattern = (
|
196 |
-
r"<s>\s*\[INST\](?P<inst>[\s\S]*?)\[/INST\](?P<answer>[\s\S]*?)</s>"
|
197 |
-
)
|
198 |
-
pair_matches = re.finditer(pair_pattern, self.merged_str, re.MULTILINE)
|
199 |
-
pair_matches_list = list(pair_matches)
|
200 |
-
|
201 |
-
self.messages = self.convert_pair_matches_to_messages(pair_matches_list)
|
202 |
-
|
203 |
-
inst_pattern = r"\[INST\](?P<inst>[\s\S]*?)\[/INST\]"
|
204 |
-
inst_matches = re.finditer(inst_pattern, self.merged_str, re.MULTILINE)
|
205 |
-
inst_matches_list = list(inst_matches)
|
206 |
-
|
207 |
-
self.append_last_instruction_to_messages(
|
208 |
-
inst_matches_list, pair_matches_list
|
209 |
-
)
|
210 |
-
elif self.model in ["nous-mixtral-8x7b"]:
|
211 |
-
# https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO#prompt-format
|
212 |
-
# message_pattern = r"<\|im_start\|>(?P<role>system|user|assistant)[\s\n]*(?P<content>[\s\S]*?)<\|im_end\|>"
|
213 |
-
message_pattern = r"<\|im_start\|>(?P<role>system|user|assistant)[\s\n]*(?P<content>[\s\S]*?)<\|im_end\|>"
|
214 |
-
message_matches = re.finditer(
|
215 |
-
message_pattern, self.merged_str, flags=re.MULTILINE | re.IGNORECASE
|
216 |
-
)
|
217 |
-
message_matches_list = list(message_matches)
|
218 |
-
logger.note(f"message_matches_list: {message_matches_list}")
|
219 |
-
for match in message_matches_list:
|
220 |
-
role = match.group("role")
|
221 |
-
content = match.group("content")
|
222 |
-
self.messages.append({"role": role, "content": content.strip()})
|
223 |
-
elif self.model in ["openchat-3.5"]:
|
224 |
-
pair_pattern = r"GPT4 Correct User:(?P<inst>[\s\S]*?)<\|end_of_turn\|>\s*GPT4 Correct Assistant:(?P<answer>[\s\S]*?)<\|end_of_turn\|>"
|
225 |
-
pair_matches = re.finditer(
|
226 |
-
pair_pattern, self.merged_str, flags=re.MULTILINE | re.IGNORECASE
|
227 |
-
)
|
228 |
-
pair_matches_list = list(pair_matches)
|
229 |
-
self.messages = self.convert_pair_matches_to_messages(pair_matches_list)
|
230 |
-
inst_pattern = r"GPT4 Correct User:(?P<inst>[\s\S]*?)<\|end_of_turn\|>"
|
231 |
-
inst_matches = re.finditer(
|
232 |
-
inst_pattern, self.merged_str, flags=re.MULTILINE | re.IGNORECASE
|
233 |
-
)
|
234 |
-
inst_matches_list = list(inst_matches)
|
235 |
-
self.append_last_instruction_to_messages(
|
236 |
-
inst_matches_list, pair_matches_list
|
237 |
-
)
|
238 |
-
# https://huggingface.co/google/gemma-7b-it#chat-template
|
239 |
-
elif self.model in ["gemma-7b"]:
|
240 |
-
pair_pattern = r"<start_of_turn>user[\s\n]*(?P<inst>[\s\S]*?)<end_of_turn>[\s\n]*<start_of_turn>model(?P<answer>[\s\S]*?)<end_of_turn>"
|
241 |
-
pair_matches = re.finditer(
|
242 |
-
pair_pattern, self.merged_str, flags=re.MULTILINE | re.IGNORECASE
|
243 |
-
)
|
244 |
-
pair_matches_list = list(pair_matches)
|
245 |
-
self.messages = self.convert_pair_matches_to_messages(pair_matches_list)
|
246 |
-
inst_pattern = r"<start_of_turn>user\n(?P<inst>[\s\S]*?)<end_of_turn>"
|
247 |
-
inst_matches = re.finditer(
|
248 |
-
inst_pattern, self.merged_str, flags=re.MULTILINE | re.IGNORECASE
|
249 |
-
)
|
250 |
-
inst_matches_list = list(inst_matches)
|
251 |
-
self.append_last_instruction_to_messages(
|
252 |
-
inst_matches_list, pair_matches_list
|
253 |
-
)
|
254 |
-
else:
|
255 |
-
self.messages = [
|
256 |
-
{
|
257 |
-
"role": "user",
|
258 |
-
"content": self.merged_str,
|
259 |
-
}
|
260 |
-
]
|
261 |
-
|
262 |
-
return self.messages
|
263 |
-
|
264 |
|
265 |
if __name__ == "__main__":
|
266 |
# model = "mixtral-8x7b"
|
267 |
# model = "nous-mixtral-8x7b"
|
268 |
-
model = "gemma-7b"
|
|
|
269 |
composer = MessageComposer(model)
|
270 |
messages = [
|
271 |
{
|
@@ -287,7 +186,5 @@ if __name__ == "__main__":
|
|
287 |
merged_str = composer.merge(messages)
|
288 |
logger.note("merged_str:")
|
289 |
logger.mesg(merged_str)
|
290 |
-
|
291 |
-
|
292 |
-
# logger.note("merged merged_str:")
|
293 |
-
# logger.mesg(composer.merge(composer.split(merged_str)))
|
|
|
1 |
import re
|
2 |
from pprint import pprint
|
3 |
+
|
4 |
+
from transformers import AutoTokenizer
|
5 |
+
|
6 |
+
from constants.models import AVAILABLE_MODELS, MODEL_MAP
|
7 |
from utils.logger import logger
|
8 |
|
9 |
|
10 |
class MessageComposer:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
def __init__(self, model: str = None):
|
12 |
+
if model in AVAILABLE_MODELS:
|
13 |
self.model = model
|
14 |
else:
|
15 |
self.model = "mixtral-8x7b"
|
16 |
+
self.model_fullname = MODEL_MAP[self.model]
|
17 |
self.system_roles = ["system"]
|
18 |
self.inst_roles = ["user", "system", "inst"]
|
19 |
self.answer_roles = ["assistant", "bot", "answer", "model"]
|
|
|
47 |
return concat_messages
|
48 |
|
49 |
def merge(self, messages) -> str:
|
50 |
+
# Templates for Chat Models
|
51 |
+
# - https://huggingface.co/docs/transformers/main/en/chat_templating
|
52 |
+
# - https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1#instruction-format
|
53 |
+
# - https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO#prompt-format
|
54 |
+
# - https://huggingface.co/openchat/openchat-3.5-0106
|
55 |
+
# - https://huggingface.co/google/gemma-7b-it#chat-template
|
56 |
+
|
57 |
# Mistral and Mixtral:
|
58 |
# <s> [INST] Instruction [/INST] Model answer </s> [INST] Follow-up instruction [/INST]
|
59 |
|
|
|
|
|
|
|
60 |
# Nous Mixtral:
|
61 |
# <|im_start|>system
|
62 |
# You are "Hermes 2".<|im_end|>
|
|
|
64 |
# Hello, who are you?<|im_end|>
|
65 |
# <|im_start|>assistant
|
66 |
|
67 |
+
# OpenChat:
|
68 |
+
# GPT4 Correct User: Hello<|end_of_turn|>GPT4 Correct Assistant: Hi<|end_of_turn|>GPT4 Correct User: How are you today?<|end_of_turn|>GPT4 Correct Assistant:
|
69 |
+
|
70 |
# Google Gemma-it
|
71 |
# <start_of_turn>user
|
72 |
# How does the brain work?<end_of_turn>
|
|
|
148 |
)
|
149 |
self.merged_str_list.append(f"{self.start_of_turn}model\n")
|
150 |
self.merged_str = "\n".join(self.merged_str_list)
|
151 |
+
# https://huggingface.co/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO#prompt-format
|
152 |
+
# https://huggingface.co/openchat/openchat-3.5-0106
|
153 |
+
# elif self.model in ["openchat-3.5", "nous-mixtral-8x7b"]:
|
154 |
else:
|
155 |
+
tokenizer = AutoTokenizer.from_pretrained("openchat/openchat-3.5-0106")
|
156 |
+
self.merged_str = tokenizer.apply_chat_template(
|
157 |
+
messages, tokenize=False, add_generation_prompt=True
|
|
|
|
|
158 |
)
|
159 |
|
160 |
return self.merged_str
|
161 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
162 |
|
163 |
if __name__ == "__main__":
|
164 |
# model = "mixtral-8x7b"
|
165 |
# model = "nous-mixtral-8x7b"
|
166 |
+
# model = "gemma-7b"
|
167 |
+
model = "openchat-3.5"
|
168 |
composer = MessageComposer(model)
|
169 |
messages = [
|
170 |
{
|
|
|
186 |
merged_str = composer.merge(messages)
|
187 |
logger.note("merged_str:")
|
188 |
logger.mesg(merged_str)
|
189 |
+
|
190 |
+
# python -m messagers.message_composer
|
|
|
|