camel-ai 0.2.67__py3-none-any.whl → 0.2.80a2__py3-none-any.whl

camel/configs/vllm_config.py

@@ -90,6 +90,7 @@ class VLLMConfig(BaseConfig):
             most likely tokens to return at each token position, each with an
             associated log probability. `logprobs` must be set to `true` if
             this parameter is used. (default: :obj:`None`)
+        extra_body: Add additional JSON properties to the request. (default: :obj:`None`)
     """
 
     temperature: Optional[float] = None  # openai default: 1.0
@@ -105,6 +106,7 @@ class VLLMConfig(BaseConfig):
     user: Optional[str] = None
     logprobs: Optional[bool] = None
     top_logprobs: Optional[int] = None
+    extra_body: Optional[dict] = None
 
 
 VLLM_API_PARAMS = {param for param in VLLMConfig.model_fields.keys()}

camel/data_collectors/alpaca_collector.py

@@ -70,16 +70,25 @@ class AlpacaDataCollector(BaseDataCollector):
         if not history:
             raise ValueError("No data collected.")
 
-        # Validate and process history
-        if len(history) == 3 and history[0].role == "system":
-            history = history[1:]  # Ignore the system message.
-        elif len(history) != 2:
+        # Filter out system and tool-related messages
+        # Keep only user and final assistant messages
+        filtered_history = []
+        for msg in history:
+            if msg.role == "user":
+                filtered_history.append(msg)
+            elif msg.role == "assistant" and msg.message:
+                # Keep assistant messages with actual content
+                # (skip empty ones that only contain tool calls)
+                filtered_history.append(msg)
+
+        # Validate filtered history
+        if len(filtered_history) != 2:
             raise ValueError(
                 f"AlpacaDataCollector only supports one message pair, but "
-                f"got {len(history)}"
+                f"got {len(filtered_history)} after filtering tool messages"
             )
 
-        input_message, output_message = history
+        input_message, output_message = filtered_history
         instruction = (
             self.system_message.content if self.system_message else ""
         ) + str(input_message.message)

camel/datagen/self_improving_cot.py

@@ -116,7 +116,7 @@ class SelfImprovingCoTPipeline:
                 samples to be drawn using the rejection sampling
                 method, where samples are accepted or rejected based on
                 a predefined condition to achieve a desired distribution.
-                (default: :obj: `None`)
+                (default: :obj:`None`)
             evaluate_agent (Optional[ChatAgent]): The chat agent used for
                 evaluating reasoning traces. (default: :obj:`None`)
             reward_model (BaseRewardModel, optional): Model used to evaluate

camel/datasets/base_generator.py

@@ -62,6 +62,7 @@ class BaseGenerator(abc.ABC, IterableDataset):
         self._buffer = buffer
         self._data: List[DataPoint] = []
         self._batch_to_save: List[DataPoint] = []
+        self._iter_position: int = 0
 
         if data_path:
             file_path = Path(data_path)
@@ -103,9 +104,9 @@ class BaseGenerator(abc.ABC, IterableDataset):
         r"""Async iterator that yields datapoints dynamically.
 
         If a `data_path` was provided during initialization, those datapoints
-        are yielded first. When self._data is empty, 20 new datapoints
-        are generated. Every 100 yields, the batch is appended to the
-        JSONL file or discarded if `cache` is None.
+        are yielded first. When self._iter_position reaches the end of _data,
+        new datapoints are generated. Every 100 yields, the batch is appended
+        to the JSONL file or discarded if `cache` is None.
 
         Yields:
             DataPoint: A single datapoint.
@@ -113,9 +114,10 @@ class BaseGenerator(abc.ABC, IterableDataset):
 
         async def generator():
             while True:
-                if not self._data:
+                if self._iter_position >= len(self._data):
                     await self.generate_new(self._buffer)
-                datapoint = self._data.pop(0)
+                datapoint = self._data[self._iter_position]
+                self._iter_position += 1
                 yield datapoint
                 self._batch_to_save.append(datapoint)
                 if len(self._batch_to_save) == 100:
@@ -132,9 +134,9 @@ class BaseGenerator(abc.ABC, IterableDataset):
         r"""Synchronous iterator for PyTorch IterableDataset compatibility.
 
         If a `data_path` was provided during initialization, those datapoints
-        are yielded first. When self._data is empty, 20 new datapoints
-        are generated. Every 100 yields, the batch is appended to the
-        JSONL file or discarded if `cache` is None.
+        are yielded first. When self._iter_position reaches the end of _data,
+        new datapoints are generated. Every 100 yields, the batch is appended
+        to the JSONL file or discarded if `cache` is None.
 
         Yields:
             DataPoint: A single datapoint.
@@ -150,9 +152,10 @@ class BaseGenerator(abc.ABC, IterableDataset):
                 raise
 
         while True:
-            if not self._data:
+            if self._iter_position >= len(self._data):
                 asyncio.run(self.generate_new(self._buffer))
-            datapoint = self._data.pop(0)
+            datapoint = self._data[self._iter_position]
+            self._iter_position += 1
             yield datapoint
             self._batch_to_save.append(datapoint)
             if len(self._batch_to_save) == 100:
@@ -248,6 +251,7 @@ class BaseGenerator(abc.ABC, IterableDataset):
 
         self.save_to_jsonl(file_path)
         self._data = []
+        self._iter_position = 0
         logger.info(f"Data flushed to {file_path} and cleared from the memory")
 
     def _init_from_jsonl(self, file_path: Path) -> List[Dict[str, Any]]:
@@ -290,3 +294,28 @@ class BaseGenerator(abc.ABC, IterableDataset):
             f"Successfully loaded {len(raw_data)} items from {file_path}"
         )
         return raw_data
+
+    def __getitem__(self, index: int) -> DataPoint:
+        r"""Get a datapoint by index without removing the datapoint from _data.
+
+        Args:
+            index (int): Index of the datapoint to retrieve.
+
+        Returns:
+            DataPoint: The datapoint at the specified index.
+
+        Raises:
+            IndexError: If the index is out of range.
+        """
+        if index < 0 or index >= len(self._data):
+            raise IndexError(f"Index {index} is out of range")
+
+        return self._data[index]
+
+    def __len__(self) -> int:
+        r"""Get the number of datapoints in the dataset.
+
+        Returns:
+            int: The number of datapoints.
+        """
+        return len(self._data)

camel/environments/__init__.py

@@ -13,6 +13,13 @@
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 from .models import Action, Environment, Observation, StepResult
 from .multi_step import MultiStepEnv
+from .rlcards_env import (
+    ActionExtractor,
+    BlackjackEnv,
+    DoudizhuEnv,
+    LeducHoldemEnv,
+    RLCardsEnv,
+)
 from .single_step import SingleStepEnv
 from .tic_tac_toe import Opponent, TicTacToeEnv
 
@@ -25,4 +32,9 @@ __all__ = [
     "StepResult",
     "TicTacToeEnv",
     "Opponent",
+    "RLCardsEnv",
+    "BlackjackEnv",
+    "LeducHoldemEnv",
+    "ActionExtractor",
+    "DoudizhuEnv",
 ]