Upload standard.py with huggingface_hub
Browse files- standard.py +79 -5
standard.py
CHANGED
@@ -1,10 +1,11 @@
|
|
1 |
from typing import List
|
2 |
|
3 |
from .card import TaskCard
|
4 |
-
from .dataclass import InternalField
|
5 |
from .formats import ICLFormat
|
6 |
from .instructions import Instruction
|
7 |
from .operator import SourceSequntialOperator, StreamingOperator
|
|
|
8 |
from .recipe import Recipe
|
9 |
from .renderers import StandardRenderer
|
10 |
from .schema import ToUnitxtGroup
|
@@ -12,14 +13,24 @@ from .splitters import Sampler, SeparateSplit, SpreadSplit
|
|
12 |
from .templates import Template
|
13 |
|
14 |
|
15 |
-
class
|
16 |
card: TaskCard
|
17 |
template: Template = None
|
18 |
instruction: Instruction = None
|
19 |
format: ICLFormat = ICLFormat()
|
20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
demos_pool_size: int = None
|
22 |
-
num_demos: int =
|
23 |
|
24 |
demos_pool_name: str = "demos_pool"
|
25 |
demos_taken_from: str = "train"
|
@@ -28,6 +39,18 @@ class StandardRecipe(Recipe, SourceSequntialOperator):
|
|
28 |
|
29 |
steps: List[StreamingOperator] = InternalField(default_factory=list)
|
30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
31 |
def prepare(self):
|
32 |
self.steps = [
|
33 |
self.card.loader,
|
@@ -47,7 +70,7 @@ class StandardRecipe(Recipe, SourceSequntialOperator):
|
|
47 |
)
|
48 |
)
|
49 |
|
50 |
-
if self.num_demos
|
51 |
sampler = self.card.sampler
|
52 |
|
53 |
if self.sampler is not None:
|
@@ -63,6 +86,15 @@ class StandardRecipe(Recipe, SourceSequntialOperator):
|
|
63 |
)
|
64 |
)
|
65 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
render = StandardRenderer(
|
67 |
instruction=self.instruction,
|
68 |
template=self.template,
|
@@ -83,7 +115,7 @@ class StandardRecipe(Recipe, SourceSequntialOperator):
|
|
83 |
)
|
84 |
|
85 |
|
86 |
-
class StandardRecipeWithIndexes(
|
87 |
instruction_card_index: int = None
|
88 |
template_card_index: int = None
|
89 |
|
@@ -101,3 +133,45 @@ class StandardRecipeWithIndexes(StandardRecipe):
|
|
101 |
self.instruction = self.card.instructions[int(self.instruction_card_index)]
|
102 |
|
103 |
super().prepare()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from typing import List
|
2 |
|
3 |
from .card import TaskCard
|
4 |
+
from .dataclass import InternalField, OptionalField
|
5 |
from .formats import ICLFormat
|
6 |
from .instructions import Instruction
|
7 |
from .operator import SourceSequntialOperator, StreamingOperator
|
8 |
+
from .operators import StreamRefiner
|
9 |
from .recipe import Recipe
|
10 |
from .renderers import StandardRenderer
|
11 |
from .schema import ToUnitxtGroup
|
|
|
13 |
from .templates import Template
|
14 |
|
15 |
|
16 |
+
class BaseRecipe(Recipe, SourceSequntialOperator):
|
17 |
card: TaskCard
|
18 |
template: Template = None
|
19 |
instruction: Instruction = None
|
20 |
format: ICLFormat = ICLFormat()
|
21 |
|
22 |
+
max_train_instances: int = None
|
23 |
+
max_validation_instances: int = None
|
24 |
+
max_test_instances: int = None
|
25 |
+
|
26 |
+
train_refiner: StreamRefiner = OptionalField(default_factory=lambda: StreamRefiner(apply_to_streams=["train"]))
|
27 |
+
validation_refiner: StreamRefiner = OptionalField(
|
28 |
+
default_factory=lambda: StreamRefiner(apply_to_streams=["validation"])
|
29 |
+
)
|
30 |
+
test_refiner: StreamRefiner = OptionalField(default_factory=lambda: StreamRefiner(apply_to_streams=["test"]))
|
31 |
+
|
32 |
demos_pool_size: int = None
|
33 |
+
num_demos: int = 0
|
34 |
|
35 |
demos_pool_name: str = "demos_pool"
|
36 |
demos_taken_from: str = "train"
|
|
|
39 |
|
40 |
steps: List[StreamingOperator] = InternalField(default_factory=list)
|
41 |
|
42 |
+
def verify(self):
|
43 |
+
super().verify()
|
44 |
+
if self.num_demos > 0:
|
45 |
+
if self.demos_pool_size is None or self.demos_pool_size < 1:
|
46 |
+
raise ValueError(
|
47 |
+
"When using demonstrations both num_demos and demos_pool_size should be assigned with postive integers."
|
48 |
+
)
|
49 |
+
if self.demos_pool_size < self.num_demos:
|
50 |
+
raise ValueError(
|
51 |
+
f"demos_pool_size must be bigger than num_demos={self.num_demos}, Got demos_pool_size={self.demos_pool_size}"
|
52 |
+
)
|
53 |
+
|
54 |
def prepare(self):
|
55 |
self.steps = [
|
56 |
self.card.loader,
|
|
|
70 |
)
|
71 |
)
|
72 |
|
73 |
+
if self.num_demos > 0:
|
74 |
sampler = self.card.sampler
|
75 |
|
76 |
if self.sampler is not None:
|
|
|
86 |
)
|
87 |
)
|
88 |
|
89 |
+
self.train_refiner.max_instances = self.max_train_instances
|
90 |
+
self.steps.append(self.train_refiner)
|
91 |
+
|
92 |
+
self.validation_refiner.max_instances = self.max_validation_instances
|
93 |
+
self.steps.append(self.validation_refiner)
|
94 |
+
|
95 |
+
self.test_refiner.max_instances = self.max_test_instances
|
96 |
+
self.steps.append(self.test_refiner)
|
97 |
+
|
98 |
render = StandardRenderer(
|
99 |
instruction=self.instruction,
|
100 |
template=self.template,
|
|
|
115 |
)
|
116 |
|
117 |
|
118 |
+
class StandardRecipeWithIndexes(BaseRecipe):
|
119 |
instruction_card_index: int = None
|
120 |
template_card_index: int = None
|
121 |
|
|
|
133 |
self.instruction = self.card.instructions[int(self.instruction_card_index)]
|
134 |
|
135 |
super().prepare()
|
136 |
+
|
137 |
+
|
138 |
+
class StandardRecipe(StandardRecipeWithIndexes):
|
139 |
+
"""
|
140 |
+
This class represents a standard recipe for data processing and preperation.
|
141 |
+
This class can be used to prepare a recipe
|
142 |
+
with all necessary steps, refiners and renderers included. It allows to set various
|
143 |
+
parameters and steps in a sequential manner for preparing the recipe.
|
144 |
+
|
145 |
+
Attributes:
|
146 |
+
card (TaskCard): TaskCard object associated with the recipe.
|
147 |
+
template (Template, optional): Template object to be used for the recipe.
|
148 |
+
instruction (Instruction, optional): Instruction object to be used for the recipe.
|
149 |
+
format (ICLFormat, optional): ICLFormat object to be used for the recipe.
|
150 |
+
train_refiner (StreamRefiner, optional): Train refiner to be used in the recipe.
|
151 |
+
max_train_instances (int, optional): Maximum training instances for the refiner.
|
152 |
+
validation_refiner (StreamRefiner, optional): Validation refiner to be used in the recipe.
|
153 |
+
max_validation_instances (int, optional): Maximum validation instances for the refiner.
|
154 |
+
test_refiner (StreamRefiner, optional): Test refiner to be used in the recipe.
|
155 |
+
max_test_instances (int, optional): Maximum test instances for the refiner.
|
156 |
+
demos_pool_size (int, optional): Size of the demos pool.
|
157 |
+
num_demos (int, optional): Number of demos to be used.
|
158 |
+
demos_pool_name (str, optional): Name of the demos pool. Default is "demos_pool".
|
159 |
+
demos_taken_from (str, optional): Specifies from where the demos are taken. Default is "train".
|
160 |
+
demos_field (str, optional): Field name for demos. Default is "demos".
|
161 |
+
sampler (Sampler, optional): Sampler object to be used in the recipe.
|
162 |
+
steps (List[StreamingOperator], optional): List of StreamingOperator objects to be used in the recipe.
|
163 |
+
instruction_card_index (int, optional): Index of instruction card to be used
|
164 |
+
for preparing the recipe.
|
165 |
+
template_card_index (int, optional): Index of template card to be used for
|
166 |
+
preparing the recipe.
|
167 |
+
|
168 |
+
Methods:
|
169 |
+
prepare(): This overridden method is used for preparing the recipe
|
170 |
+
by arranging all the steps, refiners, and renderers in a sequential manner.
|
171 |
+
|
172 |
+
Raises:
|
173 |
+
AssertionError: If both template and template_card_index, or instruction and instruction_card_index
|
174 |
+
are specified at the same time.
|
175 |
+
"""
|
176 |
+
|
177 |
+
pass
|