Upload templates.py with huggingface_hub
Browse files- templates.py +203 -0
templates.py
ADDED
@@ -0,0 +1,203 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from .text_utils import split_words
|
2 |
+
from .artifact import Artifact
|
3 |
+
from .operator import StreamInstanceOperator, InstanceOperatorWithGlobalAccess
|
4 |
+
from .instructions import Instruction
|
5 |
+
|
6 |
+
import random
|
7 |
+
from typing import Dict, Any, List
|
8 |
+
from abc import ABC, abstractmethod
|
9 |
+
|
10 |
+
|
11 |
+
class Renderer(ABC):
|
12 |
+
@abstractmethod
|
13 |
+
def get_postprocessors(self) -> List[str]:
|
14 |
+
pass
|
15 |
+
|
16 |
+
|
17 |
+
class Template(Artifact):
|
18 |
+
@abstractmethod
|
19 |
+
def process_inputs(self, inputs: Dict[str, object]) -> Dict[str, object]:
|
20 |
+
pass
|
21 |
+
|
22 |
+
@abstractmethod
|
23 |
+
def process_outputs(self, outputs: Dict[str, object]) -> Dict[str, object]:
|
24 |
+
pass
|
25 |
+
|
26 |
+
@abstractmethod
|
27 |
+
def get_postprocessors(self) -> List[str]:
|
28 |
+
pass
|
29 |
+
|
30 |
+
|
31 |
+
class RenderFormatTemplate(Renderer, StreamInstanceOperator):
|
32 |
+
template: Template = None
|
33 |
+
random_reference: bool = False
|
34 |
+
|
35 |
+
def verify(self):
|
36 |
+
assert isinstance(self.template, Template), "Template must be an instance of Template"
|
37 |
+
assert self.template is not None, "Template must be specified"
|
38 |
+
|
39 |
+
def process(self, instance: Dict[str, Any], stream_name: str = None) -> Dict[str, Any]:
|
40 |
+
return self.render(instance)
|
41 |
+
|
42 |
+
def render(self, instance: Dict[str, Any]) -> Dict[str, Any]:
|
43 |
+
inputs = instance.pop("inputs")
|
44 |
+
outputs = instance.pop("outputs")
|
45 |
+
|
46 |
+
source = self.template.process_inputs(inputs)
|
47 |
+
|
48 |
+
key, targets = next(iter(outputs.items()))
|
49 |
+
if not isinstance(targets, list):
|
50 |
+
targets = [targets]
|
51 |
+
|
52 |
+
references = [self.template.process_outputs({key: target}) for target in targets]
|
53 |
+
|
54 |
+
if self.random_reference:
|
55 |
+
target = random.choice(references)
|
56 |
+
else:
|
57 |
+
if len(references) == 0:
|
58 |
+
raise ValueError("No references found")
|
59 |
+
target = references[0] # what
|
60 |
+
|
61 |
+
return {
|
62 |
+
**instance,
|
63 |
+
"source": source,
|
64 |
+
"target": target,
|
65 |
+
"references": references,
|
66 |
+
}
|
67 |
+
|
68 |
+
def get_postprocessors(self) -> List[str]:
|
69 |
+
return self.template.get_postprocessors()
|
70 |
+
|
71 |
+
|
72 |
+
class RenderAutoFormatTemplate(RenderFormatTemplate):
|
73 |
+
def prepare(self):
|
74 |
+
if self.template is None:
|
75 |
+
self.template = AutoInputOutputTemplate()
|
76 |
+
elif isinstance(self.template, InputOutputTemplate):
|
77 |
+
self.template = AutoInputOutputTemplate(
|
78 |
+
input_format=self.template.input_format,
|
79 |
+
output_format=self.template.output_format,
|
80 |
+
)
|
81 |
+
else:
|
82 |
+
raise ValueError(
|
83 |
+
f"Template must be an instance of InputOutputTemplate or AutoInputOutputTemplate, got {type(self.template)}"
|
84 |
+
)
|
85 |
+
|
86 |
+
def render(self, instance: Dict[str, object]) -> Dict[str, object]:
|
87 |
+
if not self.template.is_complete():
|
88 |
+
self.template.infer_missing(instance["inputs"], instance["outputs"])
|
89 |
+
|
90 |
+
inputs = {key: value for key, value in instance["inputs"].items()}
|
91 |
+
|
92 |
+
return super().render({**instance, "inputs": inputs})
|
93 |
+
|
94 |
+
|
95 |
+
class CharacterSizeLimiter(Artifact):
|
96 |
+
limit: int = 1000
|
97 |
+
|
98 |
+
def check(self, text: str) -> bool:
|
99 |
+
return len(text) <= self.limit
|
100 |
+
|
101 |
+
|
102 |
+
class RenderTemplatedICL(RenderAutoFormatTemplate):
|
103 |
+
instruction: Instruction = None
|
104 |
+
input_prefix: str = "Input: "
|
105 |
+
output_prefix: str = "Output: "
|
106 |
+
instruction_prefix: str = ""
|
107 |
+
demos_field: str = None
|
108 |
+
size_limiter: Artifact = None
|
109 |
+
input_output_separator: str = "\n"
|
110 |
+
demo_separator: str = "\n\n"
|
111 |
+
demos_cache = None
|
112 |
+
|
113 |
+
def verify(self):
|
114 |
+
assert self.demos_cache is None
|
115 |
+
|
116 |
+
def render(self, instance: Dict[str, object]) -> Dict[str, object]:
|
117 |
+
if self.demos_cache is None:
|
118 |
+
self.demos_cache = instance.pop(self.demos_field, [])
|
119 |
+
else:
|
120 |
+
instance.pop(self.demos_field, None)
|
121 |
+
|
122 |
+
source = ""
|
123 |
+
|
124 |
+
example = super().render(instance)
|
125 |
+
|
126 |
+
input_str = self.input_prefix + example["source"] + self.input_output_separator + self.output_prefix
|
127 |
+
|
128 |
+
if self.instruction is not None:
|
129 |
+
source += self.instruction_prefix + self.instruction() + self.demo_separator
|
130 |
+
|
131 |
+
for demo_instance in self.demos_cache:
|
132 |
+
demo_example = super().render(demo_instance)
|
133 |
+
demo_str = (
|
134 |
+
self.input_prefix
|
135 |
+
+ demo_example["source"]
|
136 |
+
+ self.input_output_separator
|
137 |
+
+ self.output_prefix
|
138 |
+
+ demo_example["target"]
|
139 |
+
+ self.demo_separator
|
140 |
+
)
|
141 |
+
|
142 |
+
if self.size_limiter is not None:
|
143 |
+
if not self.size_limiter.check(source + demo_str + input_str + example["target"]):
|
144 |
+
continue
|
145 |
+
|
146 |
+
source += demo_str
|
147 |
+
|
148 |
+
source += input_str
|
149 |
+
|
150 |
+
return {
|
151 |
+
**example,
|
152 |
+
"source": source,
|
153 |
+
}
|
154 |
+
|
155 |
+
|
156 |
+
class InputOutputTemplate(Template):
|
157 |
+
input_format: str = None
|
158 |
+
output_format: str = None
|
159 |
+
|
160 |
+
def process_inputs(self, inputs: Dict[str, object]) -> Dict[str, object]:
|
161 |
+
return self.input_format.format(**inputs)
|
162 |
+
|
163 |
+
def process_outputs(self, outputs: Dict[str, object]) -> Dict[str, object]:
|
164 |
+
return self.output_format.format(**outputs)
|
165 |
+
|
166 |
+
def get_postprocessors(self) -> List[str]:
|
167 |
+
return ["to_string"]
|
168 |
+
|
169 |
+
|
170 |
+
class AutoInputOutputTemplate(InputOutputTemplate):
|
171 |
+
def infer_input_format(self, inputs):
|
172 |
+
input_format = ""
|
173 |
+
for key in inputs.keys():
|
174 |
+
name = " ".join(word.lower().capitalize() for word in split_words(key) if word != " ")
|
175 |
+
input_format += name + ": " + "{" + key + "}" + "\n"
|
176 |
+
self.input_format = input_format
|
177 |
+
|
178 |
+
def infer_output_format(self, outputs):
|
179 |
+
self.output_format = "{" + next(iter(outputs.keys())) + "}"
|
180 |
+
|
181 |
+
def infer_missing(self, inputs, outputs):
|
182 |
+
if self.input_format is None:
|
183 |
+
self.infer_input_format(inputs)
|
184 |
+
if self.output_format is None:
|
185 |
+
self.infer_output_format(outputs)
|
186 |
+
|
187 |
+
def is_complete(self):
|
188 |
+
return self.input_format is not None and self.output_format is not None
|
189 |
+
|
190 |
+
|
191 |
+
from .collections import ListCollection
|
192 |
+
|
193 |
+
|
194 |
+
class TemplatesList(ListCollection):
|
195 |
+
def verify(self):
|
196 |
+
for template in self.items:
|
197 |
+
assert isinstance(template, Template)
|
198 |
+
|
199 |
+
|
200 |
+
class TemplatesDict(Dict):
|
201 |
+
def verify(self):
|
202 |
+
for key, template in self.items():
|
203 |
+
assert isinstance(template, Template)
|