🚜 Use field in dataclasses

examples/datasets/hh-rlhf-helpful-base.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 import re
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -30,13 +30,20 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/hh-rlhf-helpful-base"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/hh-rlhf-helpful-base"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/hh-rlhf-helpful-base", metadata={"help": "Hugging Face repository ID to push the dataset to."}
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None, metadata={"help": "Number of workers to use for dataset processing."}
+    )
 
 
 def common_start(str1: str, str2: str) -> str:

examples/datasets/lm-human-preferences-descriptiveness.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/lm-human-preferences-descriptiveness"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/lm-human-preferences-descriptiveness"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/lm-human-preferences-descriptiveness",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 # Edge cases handling: remove the cases where all samples are the same

examples/datasets/lm-human-preferences-sentiment.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/lm-human-preferences-sentiment"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/lm-human-preferences-sentiment"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/lm-human-preferences-sentiment",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def to_prompt_completion(example, tokenizer):

examples/datasets/math_shepherd.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 import re
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from itertools import chain
 from typing import Optional
 
@@ -31,13 +31,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/math_shepherd"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/math_shepherd"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/math_shepherd",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def process_example(example):

examples/datasets/prm800k.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/prm800k"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/prm800k"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/prm800k",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def process_example(example):

examples/datasets/rlaif-v.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import features, load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/rlaif-v"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/rlaif-v"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/rlaif-v",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def to_conversational(example):

examples/datasets/tldr.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/tldr"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/tldr"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/tldr",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def to_prompt_completion(example):

examples/datasets/tldr_preference.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/tldr-preference"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/tldr-preference"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/tldr-preference",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def to_preference(example):

examples/datasets/ultrafeedback-prompt.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from dataclasses import dataclass
+from dataclasses import dataclass, field
 from typing import Optional
 
 from datasets import load_dataset
@@ -29,13 +29,22 @@ class ScriptArguments:
             Whether to push the dataset to the Hugging Face Hub.
         repo_id (`str`, *optional*, defaults to `"trl-lib/ultrafeedback-prompt"`):
             Hugging Face repository ID to push the dataset to.
-        dataset_num_proc (`Optional[int]`, *optional*, defaults to `None`):
+        dataset_num_proc (`int` or `None`, *optional*, defaults to `None`):
             Number of workers to use for dataset processing.
     """
 
-    push_to_hub: bool = False
-    repo_id: str = "trl-lib/ultrafeedback-prompt"
-    dataset_num_proc: Optional[int] = None
+    push_to_hub: bool = field(
+        default=False,
+        metadata={"help": "Whether to push the dataset to the Hugging Face Hub."},
+    )
+    repo_id: str = field(
+        default="trl-lib/ultrafeedback-prompt",
+        metadata={"help": "Hugging Face repository ID to push the dataset to."},
+    )
+    dataset_num_proc: Optional[int] = field(
+        default=None,
+        metadata={"help": "Number of workers to use for dataset processing."},
+    )
 
 
 def to_unpaired_preference(example):