Upload templates.py with huggingface_hub
Browse files- templates.py +92 -75
templates.py
CHANGED
|
@@ -9,6 +9,15 @@ from .random_utils import new_random_generator
|
|
| 9 |
from .type_utils import isoftype
|
| 10 |
|
| 11 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
class Template(StreamInstanceOperator):
|
| 13 |
"""The role of template is to take the fields of every instance and verbalize it.
|
| 14 |
|
|
@@ -26,8 +35,18 @@ class Template(StreamInstanceOperator):
|
|
| 26 |
postprocessors: List[str] = NonPositionalField(
|
| 27 |
default_factory=lambda: ["processors.to_string_stripped"]
|
| 28 |
)
|
| 29 |
-
instruction: str = NonPositionalField(
|
| 30 |
-
target_prefix: str = NonPositionalField(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
def process(
|
| 33 |
self, instance: Dict[str, Any], stream_name: Optional[str] = None
|
|
@@ -43,7 +62,12 @@ class Template(StreamInstanceOperator):
|
|
| 43 |
inputs = instance.get("inputs")
|
| 44 |
outputs = instance.get("outputs")
|
| 45 |
|
| 46 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
target, references = self.outputs_to_target_and_references(outputs)
|
| 48 |
|
| 49 |
return {
|
|
@@ -52,13 +76,17 @@ class Template(StreamInstanceOperator):
|
|
| 52 |
"target": target,
|
| 53 |
"references": references,
|
| 54 |
"instruction": instruction,
|
| 55 |
-
"target_prefix":
|
| 56 |
}
|
| 57 |
|
| 58 |
@abstractmethod
|
| 59 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 60 |
pass
|
| 61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
@abstractmethod
|
| 63 |
def outputs_to_target_and_references(
|
| 64 |
self, outputs: Dict[str, object]
|
|
@@ -68,6 +96,24 @@ class Template(StreamInstanceOperator):
|
|
| 68 |
def get_postprocessors(self) -> List[str]:
|
| 69 |
return self.postprocessors
|
| 70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
class InputOutputTemplate(Template):
|
| 73 |
"""Generate field 'source' from fields designated as input, and fields 'target' and 'references' from fields designated as output, of the processed instance.
|
|
@@ -78,30 +124,15 @@ class InputOutputTemplate(Template):
|
|
| 78 |
input_format: str = None
|
| 79 |
output_format: str = None
|
| 80 |
|
| 81 |
-
def process_template(self, template: str, data: Dict[str, object]) -> str:
|
| 82 |
-
data = {k: ", ".join(v) if isinstance(v, list) else v for k, v in data.items()}
|
| 83 |
-
return template.format(**data)
|
| 84 |
-
|
| 85 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
formatted.append(self.process_template(formatting, inputs))
|
| 90 |
-
except KeyError as e:
|
| 91 |
-
raise KeyError(
|
| 92 |
-
f"Available inputs are {list(inputs.keys())} but input format requires a different ones: '{formatting}'"
|
| 93 |
-
) from e
|
| 94 |
-
|
| 95 |
-
return tuple(formatted)
|
| 96 |
|
| 97 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 98 |
-
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
raise KeyError(
|
| 102 |
-
f"Available outputs are {outputs.keys()} but output format requires a different one: {self.output_format}"
|
| 103 |
-
) from e
|
| 104 |
-
|
| 105 |
references = [target]
|
| 106 |
return target, references
|
| 107 |
|
|
@@ -110,19 +141,13 @@ class InputOutputReferenceTemplate(InputOutputTemplate):
|
|
| 110 |
reference: str
|
| 111 |
|
| 112 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
| 118 |
-
|
| 119 |
-
|
| 120 |
-
output_fields[name] = result
|
| 121 |
-
except KeyError as e:
|
| 122 |
-
raise KeyError(
|
| 123 |
-
f"Available outputs are {outputs.keys()} but {name} requires a different one: {val}"
|
| 124 |
-
) from e
|
| 125 |
-
return output_fields["target"], [output_fields["reference"]]
|
| 126 |
|
| 127 |
|
| 128 |
class MultipleChoiceTemplate(Template):
|
|
@@ -135,7 +160,6 @@ class MultipleChoiceTemplate(Template):
|
|
| 135 |
choices_seperator: str = ", "
|
| 136 |
source_choice_format: str = "{choice_numeral}. {choice_text}"
|
| 137 |
target_choice_format: str = "{choice_numeral}"
|
| 138 |
-
add_numerals_as_field: str = None
|
| 139 |
enumerator: str = "capitals"
|
| 140 |
|
| 141 |
def prepare(self):
|
|
@@ -170,7 +194,7 @@ class MultipleChoiceTemplate(Template):
|
|
| 170 |
"XX",
|
| 171 |
]
|
| 172 |
|
| 173 |
-
def
|
| 174 |
choices = data[self.choices_field]
|
| 175 |
enumrated_choices = []
|
| 176 |
for i, choice in enumerate(choices):
|
|
@@ -182,22 +206,28 @@ class MultipleChoiceTemplate(Template):
|
|
| 182 |
)
|
| 183 |
return enumrated_choices
|
| 184 |
|
| 185 |
-
def
|
| 186 |
-
|
| 187 |
-
|
| 188 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 189 |
**inputs,
|
| 190 |
self.choices_field: self.choices_seperator.join(choices),
|
| 191 |
}
|
| 192 |
-
|
| 193 |
-
|
| 194 |
-
|
| 195 |
-
|
| 196 |
-
|
| 197 |
-
|
| 198 |
-
|
| 199 |
-
|
| 200 |
-
|
|
|
|
| 201 |
|
| 202 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 203 |
target = outputs[self.target_field]
|
|
@@ -210,7 +240,7 @@ class MultipleChoiceTemplate(Template):
|
|
| 210 |
f"MultipleChoiceTemplate could not locate textual target '{target}' in choices list: {outputs[self.choices_field]}"
|
| 211 |
) from e
|
| 212 |
|
| 213 |
-
choices = self.
|
| 214 |
|
| 215 |
try:
|
| 216 |
target = choices[target]
|
|
@@ -226,7 +256,7 @@ class MultipleChoiceTemplate(Template):
|
|
| 226 |
) -> Dict[str, Any]:
|
| 227 |
result = super().process(instance, stream_name)
|
| 228 |
if "options" not in result["outputs"]:
|
| 229 |
-
result["outputs"]["options"] = self.
|
| 230 |
instance["outputs"], self.target_choice_format
|
| 231 |
)
|
| 232 |
return result
|
|
@@ -259,18 +289,9 @@ class YesNoTemplate(Template):
|
|
| 259 |
no_answer: str = "No"
|
| 260 |
|
| 261 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 262 |
-
|
| 263 |
-
|
| 264 |
-
|
| 265 |
-
formatted = []
|
| 266 |
-
for formatting in [self.input_format, self.instruction]:
|
| 267 |
-
try:
|
| 268 |
-
formatted.append(formatting.format(**data))
|
| 269 |
-
except KeyError as e:
|
| 270 |
-
raise RuntimeError(
|
| 271 |
-
f"Available inputs are {list(inputs.keys())} but input format requires a different one: {formatting}"
|
| 272 |
-
) from e
|
| 273 |
-
return tuple(formatted)
|
| 274 |
|
| 275 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 276 |
try:
|
|
@@ -316,26 +337,22 @@ class KeyValTemplate(Template):
|
|
| 316 |
use_keys_for_outputs: bool = False
|
| 317 |
|
| 318 |
def process_dict(
|
| 319 |
-
self,
|
| 320 |
) -> str:
|
| 321 |
-
|
| 322 |
-
k: ", ".join([str(vi) for vi in v]) if isinstance(v, list) else v
|
| 323 |
-
for k, v in dic.items()
|
| 324 |
-
}
|
| 325 |
pairs = []
|
| 326 |
-
for key, val in
|
| 327 |
key_val = [key, str(val)] if use_keys else [str(val)]
|
| 328 |
pairs.append(key_val_sep.join(key_val))
|
| 329 |
return pairs_sep.join(pairs)
|
| 330 |
|
| 331 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 332 |
-
|
| 333 |
inputs,
|
| 334 |
key_val_sep=self.key_val_seperator,
|
| 335 |
pairs_sep=self.pairs_seperator,
|
| 336 |
use_keys=self.use_keys_for_inputs,
|
| 337 |
)
|
| 338 |
-
return (ret, ret)
|
| 339 |
|
| 340 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 341 |
target = self.process_dict(
|
|
|
|
| 9 |
from .type_utils import isoftype
|
| 10 |
|
| 11 |
|
| 12 |
+
class TemplateFormatKeyError(KeyError):
|
| 13 |
+
def __init__(self, template, data, data_type, format_str, format_name):
|
| 14 |
+
keys = ", ".join(data.keys())
|
| 15 |
+
super().__init__(
|
| 16 |
+
f"Available {data_type}s are [{keys}] "
|
| 17 |
+
f"but {template.__class__.__name__}.{format_name} format requires a different ones: '{format_str}'"
|
| 18 |
+
)
|
| 19 |
+
|
| 20 |
+
|
| 21 |
class Template(StreamInstanceOperator):
|
| 22 |
"""The role of template is to take the fields of every instance and verbalize it.
|
| 23 |
|
|
|
|
| 35 |
postprocessors: List[str] = NonPositionalField(
|
| 36 |
default_factory=lambda: ["processors.to_string_stripped"]
|
| 37 |
)
|
| 38 |
+
instruction: str = NonPositionalField(default="")
|
| 39 |
+
target_prefix: str = NonPositionalField(default="")
|
| 40 |
+
title_fields: List[str] = NonPositionalField(default_factory=list)
|
| 41 |
+
|
| 42 |
+
def inputs_to_instruction_and_target_prefix(self, inputs):
|
| 43 |
+
instruction = self.apply_formatting(
|
| 44 |
+
inputs, "input", self.instruction, "instruction", serialize=True
|
| 45 |
+
)
|
| 46 |
+
target_prefix = self.apply_formatting(
|
| 47 |
+
inputs, "input", self.target_prefix, "target_prefix", serialize=True
|
| 48 |
+
)
|
| 49 |
+
return instruction, target_prefix
|
| 50 |
|
| 51 |
def process(
|
| 52 |
self, instance: Dict[str, Any], stream_name: Optional[str] = None
|
|
|
|
| 62 |
inputs = instance.get("inputs")
|
| 63 |
outputs = instance.get("outputs")
|
| 64 |
|
| 65 |
+
self.set_titles(inputs)
|
| 66 |
+
|
| 67 |
+
source = self.inputs_to_source(inputs)
|
| 68 |
+
instruction, target_prefix = self.inputs_to_instruction_and_target_prefix(
|
| 69 |
+
inputs
|
| 70 |
+
)
|
| 71 |
target, references = self.outputs_to_target_and_references(outputs)
|
| 72 |
|
| 73 |
return {
|
|
|
|
| 76 |
"target": target,
|
| 77 |
"references": references,
|
| 78 |
"instruction": instruction,
|
| 79 |
+
"target_prefix": target_prefix,
|
| 80 |
}
|
| 81 |
|
| 82 |
@abstractmethod
|
| 83 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 84 |
pass
|
| 85 |
|
| 86 |
+
def set_titles(self, data):
|
| 87 |
+
for field in self.title_fields:
|
| 88 |
+
data[field] = data[field].title()
|
| 89 |
+
|
| 90 |
@abstractmethod
|
| 91 |
def outputs_to_target_and_references(
|
| 92 |
self, outputs: Dict[str, object]
|
|
|
|
| 96 |
def get_postprocessors(self) -> List[str]:
|
| 97 |
return self.postprocessors
|
| 98 |
|
| 99 |
+
def serialize_data(self, data):
|
| 100 |
+
return {
|
| 101 |
+
k: ", ".join(str(t) for t in v) if isinstance(v, list) else v
|
| 102 |
+
for k, v in data.items()
|
| 103 |
+
}
|
| 104 |
+
|
| 105 |
+
def apply_formatting(
|
| 106 |
+
self, data, data_type, format_str, format_name, serialize=False
|
| 107 |
+
) -> str:
|
| 108 |
+
if serialize:
|
| 109 |
+
data = self.serialize_data(data)
|
| 110 |
+
try:
|
| 111 |
+
return format_str.format(**data)
|
| 112 |
+
except KeyError as e:
|
| 113 |
+
raise TemplateFormatKeyError(
|
| 114 |
+
self, data, data_type, format_str, format_name
|
| 115 |
+
) from e
|
| 116 |
+
|
| 117 |
|
| 118 |
class InputOutputTemplate(Template):
|
| 119 |
"""Generate field 'source' from fields designated as input, and fields 'target' and 'references' from fields designated as output, of the processed instance.
|
|
|
|
| 124 |
input_format: str = None
|
| 125 |
output_format: str = None
|
| 126 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 127 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 128 |
+
return self.apply_formatting(
|
| 129 |
+
inputs, "input", self.input_format, "input_format", serialize=True
|
| 130 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
|
| 132 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 133 |
+
target = self.apply_formatting(
|
| 134 |
+
outputs, "output", self.output_format, "output_format", serialize=True
|
| 135 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
references = [target]
|
| 137 |
return target, references
|
| 138 |
|
|
|
|
| 141 |
reference: str
|
| 142 |
|
| 143 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 144 |
+
target = self.apply_formatting(
|
| 145 |
+
outputs, "output", self.output_format, "output_format", serialize=True
|
| 146 |
+
)
|
| 147 |
+
reference = self.apply_formatting(
|
| 148 |
+
outputs, "output", self.reference, "reference", serialize=True
|
| 149 |
+
)
|
| 150 |
+
return target, [reference]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151 |
|
| 152 |
|
| 153 |
class MultipleChoiceTemplate(Template):
|
|
|
|
| 160 |
choices_seperator: str = ", "
|
| 161 |
source_choice_format: str = "{choice_numeral}. {choice_text}"
|
| 162 |
target_choice_format: str = "{choice_numeral}"
|
|
|
|
| 163 |
enumerator: str = "capitals"
|
| 164 |
|
| 165 |
def prepare(self):
|
|
|
|
| 194 |
"XX",
|
| 195 |
]
|
| 196 |
|
| 197 |
+
def inputs_to_choices(self, data: Dict[str, object], choice_format: str) -> str:
|
| 198 |
choices = data[self.choices_field]
|
| 199 |
enumrated_choices = []
|
| 200 |
for i, choice in enumerate(choices):
|
|
|
|
| 206 |
)
|
| 207 |
return enumrated_choices
|
| 208 |
|
| 209 |
+
def inputs_to_numerals(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 210 |
+
return self.inputs_to_choices(inputs, "{choice_numeral}")
|
| 211 |
+
|
| 212 |
+
def prepare_multiple_choice_inputs(
|
| 213 |
+
self, inputs: Dict[str, object]
|
| 214 |
+
) -> Dict[str, object]:
|
| 215 |
+
choices = self.inputs_to_choices(inputs, self.source_choice_format)
|
| 216 |
+
return {
|
| 217 |
+
"numerals": self.inputs_to_numerals(inputs),
|
| 218 |
**inputs,
|
| 219 |
self.choices_field: self.choices_seperator.join(choices),
|
| 220 |
}
|
| 221 |
+
|
| 222 |
+
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 223 |
+
inputs = self.prepare_multiple_choice_inputs(inputs)
|
| 224 |
+
return self.apply_formatting(
|
| 225 |
+
inputs, "input", self.input_format, "input_format", serialize=True
|
| 226 |
+
)
|
| 227 |
+
|
| 228 |
+
def inputs_to_instruction_and_target_prefix(self, inputs):
|
| 229 |
+
inputs = self.prepare_multiple_choice_inputs(inputs)
|
| 230 |
+
return super().inputs_to_instruction_and_target_prefix(inputs)
|
| 231 |
|
| 232 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 233 |
target = outputs[self.target_field]
|
|
|
|
| 240 |
f"MultipleChoiceTemplate could not locate textual target '{target}' in choices list: {outputs[self.choices_field]}"
|
| 241 |
) from e
|
| 242 |
|
| 243 |
+
choices = self.inputs_to_choices(outputs, self.target_choice_format)
|
| 244 |
|
| 245 |
try:
|
| 246 |
target = choices[target]
|
|
|
|
| 256 |
) -> Dict[str, Any]:
|
| 257 |
result = super().process(instance, stream_name)
|
| 258 |
if "options" not in result["outputs"]:
|
| 259 |
+
result["outputs"]["options"] = self.inputs_to_choices(
|
| 260 |
instance["outputs"], self.target_choice_format
|
| 261 |
)
|
| 262 |
return result
|
|
|
|
| 289 |
no_answer: str = "No"
|
| 290 |
|
| 291 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 292 |
+
return self.apply_formatting(
|
| 293 |
+
inputs, "input", self.input_format, "input_format", serialize=True
|
| 294 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 295 |
|
| 296 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 297 |
try:
|
|
|
|
| 337 |
use_keys_for_outputs: bool = False
|
| 338 |
|
| 339 |
def process_dict(
|
| 340 |
+
self, data: Dict[str, object], key_val_sep, pairs_sep, use_keys
|
| 341 |
) -> str:
|
| 342 |
+
data = self.serialize_data(data)
|
|
|
|
|
|
|
|
|
|
| 343 |
pairs = []
|
| 344 |
+
for key, val in data.items():
|
| 345 |
key_val = [key, str(val)] if use_keys else [str(val)]
|
| 346 |
pairs.append(key_val_sep.join(key_val))
|
| 347 |
return pairs_sep.join(pairs)
|
| 348 |
|
| 349 |
def inputs_to_source(self, inputs: Dict[str, object]) -> Tuple[str, str]:
|
| 350 |
+
return self.process_dict(
|
| 351 |
inputs,
|
| 352 |
key_val_sep=self.key_val_seperator,
|
| 353 |
pairs_sep=self.pairs_seperator,
|
| 354 |
use_keys=self.use_keys_for_inputs,
|
| 355 |
)
|
|
|
|
| 356 |
|
| 357 |
def outputs_to_target_and_references(self, outputs: Dict[str, object]) -> str:
|
| 358 |
target = self.process_dict(
|