fix prompters, especially the sharegpt prompter
Browse files- src/axolotl/prompt_tokenizers.py +72 -12
- src/axolotl/prompters.py +18 -72
src/axolotl/prompt_tokenizers.py
CHANGED
@@ -1,7 +1,10 @@
|
|
1 |
import abc
|
|
|
2 |
|
3 |
from transformers import PreTrainedTokenizer
|
4 |
|
|
|
|
|
5 |
IGNORE_INDEX = -100
|
6 |
LLAMA_DEFAULT_PAD_TOKEN = "[PAD]"
|
7 |
LLAMA_DEFAULT_EOS_TOKEN = "</s>"
|
@@ -40,10 +43,10 @@ class InstructionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
|
40 |
full_prompt = self._build_full_prompt(instruction, input, response)
|
41 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
42 |
if not self.train_on_inputs:
|
43 |
-
user_prompt = self.prompter.build_prompt(
|
44 |
instruction,
|
45 |
input,
|
46 |
-
)
|
47 |
tokenized_user_prompt = self._tokenize(user_prompt, add_eos_token=False)
|
48 |
user_prompt_len = len(tokenized_user_prompt["input_ids"])
|
49 |
# TODO this could be sped up using numpy array slicing
|
@@ -54,11 +57,11 @@ class InstructionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
|
54 |
return tokenized_full_prompt
|
55 |
|
56 |
def _build_full_prompt(self, instruction, input, response):
|
57 |
-
return self.prompter.build_prompt(
|
58 |
instruction,
|
59 |
input,
|
60 |
response,
|
61 |
-
)
|
62 |
|
63 |
def _tokenize(self, prompt, add_eos_token=True):
|
64 |
result = self.tokenizer(
|
@@ -131,13 +134,13 @@ class CompletionPromptTokenizingStrategy(InstructionPromptTokenizingStrategy):
|
|
131 |
|
132 |
def tokenize_prompt(self, prompt):
|
133 |
instruction = self.parse_instruction_fields(prompt)
|
134 |
-
full_prompt = self._build_full_prompt(instruction)
|
135 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
136 |
|
137 |
return tokenized_full_prompt
|
138 |
|
139 |
-
def _build_full_prompt(self, instruction):
|
140 |
-
return self.prompter.build_prompt(instruction)
|
141 |
|
142 |
|
143 |
class ReflectionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
@@ -157,10 +160,10 @@ class ReflectionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
|
157 |
)
|
158 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
159 |
if not self.train_on_inputs:
|
160 |
-
user_prompt = self.prompter.build_prompt(
|
161 |
instruction,
|
162 |
input,
|
163 |
-
)
|
164 |
tokenized_user_prompt = self._tokenize(user_prompt, add_eos_token=False)
|
165 |
user_prompt_len = len(tokenized_user_prompt["input_ids"])
|
166 |
# TODO this could be sped up using numpy array slicing
|
@@ -171,13 +174,13 @@ class ReflectionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
|
171 |
return tokenized_full_prompt
|
172 |
|
173 |
def _build_full_prompt(self, instruction, input, output, reflection, corrected):
|
174 |
-
return self.prompter.build_prompt(
|
175 |
instruction,
|
176 |
input,
|
177 |
output,
|
178 |
reflection,
|
179 |
corrected,
|
180 |
-
)
|
181 |
|
182 |
def _tokenize(self, prompt, add_eos_token=True):
|
183 |
result = self.tokenizer(
|
@@ -212,7 +215,64 @@ class AlpacaReflectionPTStrategy(ReflectionPromptTokenizingStrategy):
|
|
212 |
|
213 |
class ShareGPTPromptTokenizingStrategy(PromptTokenizingStrategy):
|
214 |
def tokenize_prompt(self, prompt):
|
|
|
|
|
|
|
|
|
|
|
|
|
215 |
try:
|
216 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
217 |
except (KeyError, AssertionError, IndexError) as e:
|
218 |
raise InvalidDataException(str(e))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import abc
|
2 |
+
import copy
|
3 |
|
4 |
from transformers import PreTrainedTokenizer
|
5 |
|
6 |
+
from axolotl.prompters import IGNORE_TOKEN_ID
|
7 |
+
|
8 |
IGNORE_INDEX = -100
|
9 |
LLAMA_DEFAULT_PAD_TOKEN = "[PAD]"
|
10 |
LLAMA_DEFAULT_EOS_TOKEN = "</s>"
|
|
|
43 |
full_prompt = self._build_full_prompt(instruction, input, response)
|
44 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
45 |
if not self.train_on_inputs:
|
46 |
+
user_prompt = next(iter(self.prompter.build_prompt(
|
47 |
instruction,
|
48 |
input,
|
49 |
+
)))
|
50 |
tokenized_user_prompt = self._tokenize(user_prompt, add_eos_token=False)
|
51 |
user_prompt_len = len(tokenized_user_prompt["input_ids"])
|
52 |
# TODO this could be sped up using numpy array slicing
|
|
|
57 |
return tokenized_full_prompt
|
58 |
|
59 |
def _build_full_prompt(self, instruction, input, response):
|
60 |
+
return next(iter(self.prompter.build_prompt(
|
61 |
instruction,
|
62 |
input,
|
63 |
response,
|
64 |
+
)))
|
65 |
|
66 |
def _tokenize(self, prompt, add_eos_token=True):
|
67 |
result = self.tokenizer(
|
|
|
134 |
|
135 |
def tokenize_prompt(self, prompt):
|
136 |
instruction = self.parse_instruction_fields(prompt)
|
137 |
+
full_prompt = self._build_full_prompt(instruction, None, None)
|
138 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
139 |
|
140 |
return tokenized_full_prompt
|
141 |
|
142 |
+
def _build_full_prompt(self, instruction, input, response):
|
143 |
+
return next(iter(self.prompter.build_prompt(instruction)))
|
144 |
|
145 |
|
146 |
class ReflectionPromptTokenizingStrategy(PromptTokenizingStrategy):
|
|
|
160 |
)
|
161 |
tokenized_full_prompt = self._tokenize(full_prompt)
|
162 |
if not self.train_on_inputs:
|
163 |
+
user_prompt = next(iter(self.prompter.build_prompt(
|
164 |
instruction,
|
165 |
input,
|
166 |
+
)))
|
167 |
tokenized_user_prompt = self._tokenize(user_prompt, add_eos_token=False)
|
168 |
user_prompt_len = len(tokenized_user_prompt["input_ids"])
|
169 |
# TODO this could be sped up using numpy array slicing
|
|
|
174 |
return tokenized_full_prompt
|
175 |
|
176 |
def _build_full_prompt(self, instruction, input, output, reflection, corrected):
|
177 |
+
return next(iter(self.prompter.build_prompt(
|
178 |
instruction,
|
179 |
input,
|
180 |
output,
|
181 |
reflection,
|
182 |
corrected,
|
183 |
+
)))
|
184 |
|
185 |
def _tokenize(self, prompt, add_eos_token=True):
|
186 |
result = self.tokenizer(
|
|
|
215 |
|
216 |
class ShareGPTPromptTokenizingStrategy(PromptTokenizingStrategy):
|
217 |
def tokenize_prompt(self, prompt):
|
218 |
+
result = {
|
219 |
+
"input_ids": [],
|
220 |
+
"attention_mask": [],
|
221 |
+
"labels": [],
|
222 |
+
}
|
223 |
+
current_len = 0
|
224 |
try:
|
225 |
+
for i, part in enumerate(self.prompter.build_prompt(prompt["conversations"], self.tokenizer)):
|
226 |
+
if i == 0:
|
227 |
+
# this is only ever the first part, should include the bos token and the user query
|
228 |
+
res = self._tokenize(part.strip(), add_eos_token=False, strip_bos_token=False)
|
229 |
+
# everything from this is masked out from the labels
|
230 |
+
labels = [ IGNORE_TOKEN_ID ] * len(res["input_ids"])
|
231 |
+
elif i % 2 == 0:
|
232 |
+
# this is still the user query, we should
|
233 |
+
res = self._tokenize(part.strip(), add_eos_token=False, strip_bos_token=True)
|
234 |
+
# everything from this is masked out from the labels
|
235 |
+
labels = [ IGNORE_TOKEN_ID ] * len(res["input_ids"])
|
236 |
+
else:
|
237 |
+
# this should be the assistent response, should end with an eos token
|
238 |
+
res = self._tokenize(part.strip(), add_eos_token=True, strip_bos_token=True)
|
239 |
+
# not masked out from labels
|
240 |
+
labels = copy.deepcopy(res["input_ids"])
|
241 |
+
input_ids = res["input_ids"]
|
242 |
+
input_len = len(input_ids)
|
243 |
+
result["input_ids"][current_len : current_len + input_len] = input_ids
|
244 |
+
result["attention_mask"][current_len : current_len + input_len] = [
|
245 |
+
1 if x != self.tokenizer.pad_token_id else 0
|
246 |
+
for x in input_ids
|
247 |
+
]
|
248 |
+
result["labels"][current_len : current_len + input_len] = labels
|
249 |
+
current_len += input_len
|
250 |
+
return result
|
251 |
except (KeyError, AssertionError, IndexError) as e:
|
252 |
raise InvalidDataException(str(e))
|
253 |
+
|
254 |
+
def _tokenize(self, prompt, add_eos_token=True, strip_bos_token=False):
|
255 |
+
result = self.tokenizer(
|
256 |
+
prompt,
|
257 |
+
truncation=True,
|
258 |
+
max_length=self.sequence_len,
|
259 |
+
padding=False,
|
260 |
+
return_tensors=None,
|
261 |
+
)
|
262 |
+
if (
|
263 |
+
result["input_ids"][-1] != self.tokenizer.eos_token_id
|
264 |
+
and len(result["input_ids"]) < self.sequence_len
|
265 |
+
and add_eos_token
|
266 |
+
):
|
267 |
+
result["input_ids"].append(self.tokenizer.eos_token_id)
|
268 |
+
result["attention_mask"].append(1)
|
269 |
+
|
270 |
+
if (
|
271 |
+
result["input_ids"][0] == self.tokenizer.bos_token_id
|
272 |
+
and strip_bos_token
|
273 |
+
):
|
274 |
+
result["input_ids"] = result["input_ids"][1:]
|
275 |
+
result["attention_mask"] = result["attention_mask"][1:]
|
276 |
+
|
277 |
+
result["labels"] = result["input_ids"].copy()
|
278 |
+
return result
|
src/axolotl/prompters.py
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
import copy
|
2 |
import dataclasses
|
3 |
from enum import auto, Enum
|
4 |
-
from typing import List, Tuple, Any, Union
|
5 |
|
6 |
IGNORE_TOKEN_ID = -100
|
7 |
|
@@ -16,7 +16,7 @@ class AlpacaPrompter:
|
|
16 |
instruction: str,
|
17 |
input: Union[None, str] = None,
|
18 |
output: Union[None, str] = None,
|
19 |
-
) -> str:
|
20 |
# returns the full prompt from instruction and optional input
|
21 |
# if a label (=response, =output) is provided, it's also appended.
|
22 |
if input:
|
@@ -25,7 +25,7 @@ class AlpacaPrompter:
|
|
25 |
res = self.prompt_no_input.format(instruction=instruction)
|
26 |
if output:
|
27 |
res = f"{res}{output}"
|
28 |
-
|
29 |
|
30 |
def get_response(self, output: str) -> str:
|
31 |
return output.split(self.response_split)[1].strip()
|
@@ -36,8 +36,8 @@ class JeopardyPrompter(AlpacaPrompter):
|
|
36 |
|
37 |
|
38 |
class CompletionPrompter(AlpacaPrompter):
|
39 |
-
def build_prompt(self, instruction: str) -> str:
|
40 |
-
|
41 |
|
42 |
def get_response(self, output: str) -> str:
|
43 |
return output.strip()
|
@@ -64,7 +64,7 @@ class ReflectAlpacaPrompter:
|
|
64 |
output: Union[None, str] = None,
|
65 |
reflection: Union[None, str] = None,
|
66 |
corrected: Union[None, str] = None,
|
67 |
-
) -> str:
|
68 |
# returns the full prompt from instruction and optional input
|
69 |
# if a label (=response, =output) is provided, it's also appended.
|
70 |
if input:
|
@@ -76,7 +76,7 @@ class ReflectAlpacaPrompter:
|
|
76 |
output=output, reflection=reflection, corrected=corrected
|
77 |
)
|
78 |
res = f"{res}{label}"
|
79 |
-
|
80 |
|
81 |
def get_response(self, output: str) -> str:
|
82 |
return output.split(self.response_split)[1].strip()
|
@@ -103,15 +103,16 @@ class Conversation:
|
|
103 |
sep: str = "###"
|
104 |
sep2: str = None
|
105 |
|
106 |
-
def get_prompt(self):
|
107 |
seps = [self.sep, self.sep2]
|
108 |
-
|
109 |
for i, (role, message) in enumerate(self.messages):
|
110 |
if message:
|
111 |
-
|
112 |
else:
|
113 |
-
|
114 |
-
|
|
|
115 |
|
116 |
def copy(self):
|
117 |
return Conversation(
|
@@ -136,12 +137,12 @@ conv_vicuna_v1_1 = Conversation(
|
|
136 |
offset=0,
|
137 |
sep_style=SeparatorStyle.TWO,
|
138 |
sep=" ",
|
139 |
-
sep2="
|
140 |
)
|
141 |
|
142 |
|
143 |
class ShareGPTPrompter:
|
144 |
-
def build_prompt(self, source, tokenizer, sequence_len=2048):
|
145 |
# ignore the system prompt if provided
|
146 |
if source[0]["from"] == "system":
|
147 |
source.pop(0)
|
@@ -171,61 +172,6 @@ class ShareGPTPrompter:
|
|
171 |
role = roles[sentence["from"]]
|
172 |
assert role == conv.roles[j % 2]
|
173 |
conv.append_message(role, sentence["value"])
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
# Tokenize conversations
|
178 |
-
tokenized_result = tokenizer(
|
179 |
-
conversation,
|
180 |
-
truncation=True,
|
181 |
-
max_length=sequence_len, # FIXME
|
182 |
-
padding=False,
|
183 |
-
return_tensors=None,
|
184 |
-
)
|
185 |
-
target = copy.deepcopy(tokenized_result["input_ids"])
|
186 |
-
|
187 |
-
# Mask targets
|
188 |
-
sep = conv.sep + conv.roles[1] + ": "
|
189 |
-
|
190 |
-
rounds = conversation.split(conv.sep2)
|
191 |
-
rounds = [r + conv.sep2 for r in rounds]
|
192 |
-
cur_len = 1
|
193 |
-
target[0] = IGNORE_TOKEN_ID # mask out the bos
|
194 |
-
for i, rou in enumerate(rounds):
|
195 |
-
if rou == "":
|
196 |
-
break
|
197 |
-
|
198 |
-
parts = rou.split(sep)
|
199 |
-
if len(parts) != 2:
|
200 |
-
break
|
201 |
-
parts[0] += sep
|
202 |
-
round_len = (
|
203 |
-
len(tokenizer(rou)["input_ids"]) - 1
|
204 |
-
) # -1 ignores the bos_token generated for this
|
205 |
-
# we have to strip the initial part, any dangling whitespace creates an additional ghost token
|
206 |
-
instruction_len = (
|
207 |
-
len(tokenizer(parts[0].strip())["input_ids"]) - 1
|
208 |
-
) # -1 ignores the bos_token generated for this
|
209 |
-
target[cur_len : cur_len + instruction_len] = [
|
210 |
-
IGNORE_TOKEN_ID
|
211 |
-
] * instruction_len
|
212 |
-
|
213 |
-
cur_len += round_len
|
214 |
-
if cur_len >= sequence_len:
|
215 |
-
break
|
216 |
-
|
217 |
-
# Fix: Truncate the target to have the same length as input_ids
|
218 |
-
target = target[: len(tokenized_result["input_ids"])]
|
219 |
-
# target[cur_len:] = [IGNORE_TOKEN_ID] * (len(target) - cur_len)
|
220 |
-
|
221 |
-
attention_mask = [
|
222 |
-
1 if x != tokenizer.pad_token_id else 0
|
223 |
-
for x in tokenized_result["input_ids"]
|
224 |
-
]
|
225 |
-
|
226 |
-
# TODO truncate len to sequence_len
|
227 |
-
return dict(
|
228 |
-
input_ids=tokenized_result["input_ids"],
|
229 |
-
labels=target,
|
230 |
-
attention_mask=attention_mask,
|
231 |
-
)
|
|
|
1 |
import copy
|
2 |
import dataclasses
|
3 |
from enum import auto, Enum
|
4 |
+
from typing import List, Tuple, Any, Union, Generator
|
5 |
|
6 |
IGNORE_TOKEN_ID = -100
|
7 |
|
|
|
16 |
instruction: str,
|
17 |
input: Union[None, str] = None,
|
18 |
output: Union[None, str] = None,
|
19 |
+
) -> Generator[str, None, None]:
|
20 |
# returns the full prompt from instruction and optional input
|
21 |
# if a label (=response, =output) is provided, it's also appended.
|
22 |
if input:
|
|
|
25 |
res = self.prompt_no_input.format(instruction=instruction)
|
26 |
if output:
|
27 |
res = f"{res}{output}"
|
28 |
+
yield res
|
29 |
|
30 |
def get_response(self, output: str) -> str:
|
31 |
return output.split(self.response_split)[1].strip()
|
|
|
36 |
|
37 |
|
38 |
class CompletionPrompter(AlpacaPrompter):
|
39 |
+
def build_prompt(self, instruction: str, input=None, output=None) -> Generator[str, None, None]:
|
40 |
+
yield instruction
|
41 |
|
42 |
def get_response(self, output: str) -> str:
|
43 |
return output.strip()
|
|
|
64 |
output: Union[None, str] = None,
|
65 |
reflection: Union[None, str] = None,
|
66 |
corrected: Union[None, str] = None,
|
67 |
+
) -> Generator[str, None, None]:
|
68 |
# returns the full prompt from instruction and optional input
|
69 |
# if a label (=response, =output) is provided, it's also appended.
|
70 |
if input:
|
|
|
76 |
output=output, reflection=reflection, corrected=corrected
|
77 |
)
|
78 |
res = f"{res}{label}"
|
79 |
+
yield res
|
80 |
|
81 |
def get_response(self, output: str) -> str:
|
82 |
return output.split(self.response_split)[1].strip()
|
|
|
103 |
sep: str = "###"
|
104 |
sep2: str = None
|
105 |
|
106 |
+
def get_prompt(self) -> Generator[str, None, None]:
|
107 |
seps = [self.sep, self.sep2]
|
108 |
+
preamble = self.system + seps[0]
|
109 |
for i, (role, message) in enumerate(self.messages):
|
110 |
if message:
|
111 |
+
yield preamble + role + ": " + message + seps[i % 2]
|
112 |
else:
|
113 |
+
yield role + ":"
|
114 |
+
if i == 0:
|
115 |
+
preamble = ""
|
116 |
|
117 |
def copy(self):
|
118 |
return Conversation(
|
|
|
137 |
offset=0,
|
138 |
sep_style=SeparatorStyle.TWO,
|
139 |
sep=" ",
|
140 |
+
sep2=" ",
|
141 |
)
|
142 |
|
143 |
|
144 |
class ShareGPTPrompter:
|
145 |
+
def build_prompt(self, source, tokenizer, sequence_len=2048) -> Generator[str, None, None]:
|
146 |
# ignore the system prompt if provided
|
147 |
if source[0]["from"] == "system":
|
148 |
source.pop(0)
|
|
|
172 |
role = roles[sentence["from"]]
|
173 |
assert role == conv.roles[j % 2]
|
174 |
conv.append_message(role, sentence["value"])
|
175 |
+
|
176 |
+
for part in conv.get_prompt():
|
177 |
+
yield part
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|