Convert module-level optional dependency imports to lazy imports with @dependencies_required

camel/benchmarks/apibank.py

@@ -21,10 +21,6 @@
 from pathlib import Path
 from typing import Any, Dict, List, Literal, Optional
 
-import numpy as np
-from rouge import Rouge
-from tqdm import tqdm
-
 from camel.agents import ChatAgent
 from camel.benchmarks.base import BaseBenchmark
 from camel.messages import BaseMessage
@@ -138,6 +134,8 @@ def load(self, level: str, force_download: bool = False):  # type: ignore[overri
         jsonl_files = [
             f for f in os.listdir(file_path) if f.endswith('.jsonl')
         ]
+        from tqdm import tqdm
+
         for file in tqdm(jsonl_files, desc="Processing files"):
             history = []
             with open(file_path / file, 'r') as f:
@@ -240,6 +238,8 @@ def run(  # type: ignore[override, return]
         total_api_calls, correct_api_calls, rougel_scores = 0, 0, []
 
         with open(self.save_to, "w") as f:
+            from tqdm import tqdm
+
             for test in tqdm(datas, desc="Running"):
                 samples = self._data[test]
                 evaluator = Evaluator(samples)  # type: ignore[arg-type]
@@ -380,6 +380,8 @@ def run(  # type: ignore[override, return]
                 else 0,
             }
         elif dialog_test_enabled:
+            import numpy as np
+
             return {'Dialog_score': np.mean(rougel_scores)}
 
 
@@ -415,6 +417,8 @@ def agent_call(messages: List[Dict], agent: ChatAgent):
 
 def calculate_rouge_l_score(reference, hypothesis):
     r"""Calculate rouge l score between hypothesis and reference."""
+    from rouge import Rouge
+
     rouge = Rouge()
     scores = rouge.get_scores(hypothesis, reference)
     rouge_l_score = scores[0]['rouge-l']['f']

camel/benchmarks/apibench.py

@@ -18,10 +18,6 @@
 from pathlib import Path
 from typing import Any, Dict, Literal, Optional
 
-import tree_sitter_python as tspython
-from tqdm import tqdm
-from tree_sitter import Language, Parser
-
 from camel.agents import ChatAgent
 from camel.benchmarks.base import BaseBenchmark
 from camel.utils import download_github_subdirectory
@@ -278,6 +274,8 @@ def run(  # type: ignore[override]
         self._results = []
 
         with open(self.save_to, "w") as f:
+            from tqdm import tqdm
+
             for question in tqdm(datas, desc="Running"):
                 prompt = encode_question(question["text"], dataset_name)
                 try:
@@ -376,6 +374,9 @@ def get_all_sub_trees(root_node):
 
 # Parse the program into AST trees
 def ast_parse(candidate):
+    import tree_sitter_python as tspython
+    from tree_sitter import Language, Parser
+
     PY_LANGUAGE = Language(tspython.language())
     parser = Parser(PY_LANGUAGE)
 

camel/benchmarks/gaia.py

@@ -22,8 +22,6 @@
 from pathlib import Path
 from typing import Any, Dict, List, Literal, Optional, Protocol, Union
 
-from tqdm import tqdm
-
 from camel.agents import ChatAgent
 from camel.benchmarks.base import BaseBenchmark
 from camel.messages import BaseMessage
@@ -261,6 +259,8 @@ def run(  # type: ignore[override]
 
         # Process tasks
         with open(self.save_to, "w") as f:
+            from tqdm import tqdm
+
             for task in tqdm(datas, desc="Running"):
                 if not self._prepare_task(task):
                     continue

camel/benchmarks/nexus.py

@@ -22,12 +22,9 @@
 from pathlib import Path
 from typing import Any, Dict, List, Literal, Optional, Tuple, Union
 
-import pandas as pd
-from datasets import load_dataset
-from tqdm import tqdm
-
 from camel.agents import ChatAgent
 from camel.benchmarks.base import BaseBenchmark
+from camel.utils import dependencies_required
 
 logger = logging.getLogger(__name__)
 
@@ -89,6 +86,7 @@ class NexusBenchmark(BaseBenchmark):
             (default: :obj:`1`)
     """
 
+    @dependencies_required('datasets', 'pandas')
     def __init__(
         self,
         data_dir: str,
@@ -126,6 +124,7 @@ def load(self, dataset_name: str, force_download: bool = False):  # type: ignore
             dataset_name (str): Name of the specific dataset to be loaded.
             force_download (bool): Whether to force download the data.
         """
+        import pandas as pd
 
         def _load_csv_data(dataset_dir: Path) -> List:
             r"""Load datasets from CSV files."""
@@ -306,6 +305,8 @@ def run(  # type: ignore[override, return]
         # Process samples
         tools = construct_tool_descriptions(task)
         with open(self.save_to, "w") as f:
+            from tqdm import tqdm
+
             for sample in tqdm(datas, desc="Running"):
                 prompt = construct_prompt(input=sample.input, tools=tools)
                 ground_truth_call = sample.output
@@ -363,6 +364,8 @@ def run(  # type: ignore[override, return]
 def construct_tool_descriptions(dataset_name: str) -> str:
     r"""Construct tool descriptions from function definitions and
     descriptions."""
+    from datasets import load_dataset
+
     tool_dataset_mapping = {
         "NVDLibrary": "CVECPE",
         "VirusTotal": "VirusTotal",

camel/benchmarks/ragbench.py

@@ -12,15 +12,25 @@
 # limitations under the License.
 # ========= Copyright 2023-2026 @ CAMEL-AI.org. All Rights Reserved. =========
 
-from typing import Any, Callable, Dict, List, Literal, Optional, Sequence
-
-import numpy as np
-from datasets import Dataset, load_dataset
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Callable,
+    Dict,
+    List,
+    Literal,
+    Optional,
+    Sequence,
+)
+
+if TYPE_CHECKING:
+    from datasets import Dataset
 
 from camel.agents import ChatAgent
 from camel.benchmarks import BaseBenchmark
 from camel.logger import get_logger
 from camel.retrievers import AutoRetriever
+from camel.utils import dependencies_required
 
 logger = get_logger(__name__)
 
@@ -34,10 +44,10 @@ class RagasFields:
 
 
 def annotate_dataset(
-    dataset: Dataset,
+    dataset: 'Dataset',
     context_call: Optional[Callable[[Dict[str, Any]], List[str]]],
     answer_call: Optional[Callable[[Dict[str, Any]], str]],
-) -> Dataset:
+) -> 'Dataset':
     r"""Annotate the dataset by adding context and answers using the provided
     functions.
 
@@ -75,6 +85,8 @@ def rmse(
     Returns:
         Optional[float]: RMSE value, or None if inputs have different lengths.
     """
+    import numpy as np
+
     if len(input_trues) != len(input_preds):
         logger.warning("Input lengths mismatch in RMSE calculation")
         return None
@@ -104,6 +116,7 @@ def auroc(trues: Sequence[bool], preds: Sequence[float]) -> float:
     Returns:
         float: AUROC score.
     """
+    import numpy as np
     from sklearn.metrics import roc_auc_score  # type: ignore[import-untyped]
 
     eval_idx = ~np.isnan(preds)
@@ -117,7 +130,7 @@ def auroc(trues: Sequence[bool], preds: Sequence[float]) -> float:
 
 
 def ragas_calculate_metrics(
-    dataset: Dataset,
+    dataset: 'Dataset',
     pred_context_relevance_field: Optional[str],
     pred_faithfulness_field: Optional[str],
     metrics_to_evaluate: Optional[List[str]] = None,
@@ -141,6 +154,8 @@ def ragas_calculate_metrics(
     Returns:
         Dict[str, Optional[float]]: Dictionary of calculated metrics.
     """
+    import numpy as np
+
     metrics_to_evaluate = metrics_to_evaluate or [
         "context_relevancy",
         "faithfulness",
@@ -172,11 +187,11 @@ def ragas_calculate_metrics(
 
 
 def ragas_evaluate_dataset(
-    dataset: Dataset,
+    dataset: 'Dataset',
     contexts_field_name: Optional[str],
     answer_field_name: Optional[str],
     metrics_to_evaluate: Optional[List[str]] = None,
-) -> Dataset:
+) -> 'Dataset':
     r"""Evaluate the dataset using RAGAS metrics.
 
     Args:
@@ -188,6 +203,7 @@ def ragas_evaluate_dataset(
     Returns:
         Dataset: Dataset with added evaluation metrics.
     """
+    from datasets import Dataset
     from ragas import evaluate  # type: ignore[import]
     from ragas.metrics import (  # type: ignore[import]
         context_relevancy,
@@ -236,6 +252,7 @@ class RAGBenchBenchmark(BaseBenchmark):
         split (str, optional): Dataset split to use (e.g., "test").
     """
 
+    @dependencies_required('datasets')
     def __init__(
         self,
         processes: int = 1,
@@ -258,10 +275,12 @@ def __init__(
         super().__init__("ragbench", "rag_bench", "", processes)
         self.subset = subset
         self.split = split
-        self.dataset: Optional[Dataset] = None
+        self.dataset: Optional['Dataset'] = None
 
     def download(self):
         r"""Download the RAGBench dataset."""
+        from datasets import load_dataset
+
         try:
             self.dataset = load_dataset(
                 "rungalileo/ragbench", self.subset, split=self.split

camel/datagen/evol_instruct/evol_instruct.py

@@ -18,8 +18,6 @@
 from math import ceil
 from typing import Any, Dict, List, Optional, Tuple, Type, Union, cast
 
-from tqdm import tqdm
-
 from camel.agents import ChatAgent
 from camel.datagen.evol_instruct.scorer import BaseScorer, GeneralScorer
 from camel.datagen.evol_instruct.templates import EvolInstructTemplates
@@ -413,6 +411,8 @@ def _process_prompt(
             plan_chunk = evolution_plan[chunk_idx : chunk_idx + chunk_size]
 
             with ThreadPoolExecutor(max_workers=num_threads) as executor:
+                from tqdm import tqdm
+
                 chunk_results = list(
                     tqdm(
                         executor.map(_process_prompt, zip(chunk, plan_chunk)),

camel/datagen/self_instruct/filter/filter_function.py

@@ -16,8 +16,6 @@
 from abc import ABC, abstractmethod
 from typing import List
 
-from rouge import Rouge
-
 from camel.models.reward import BaseRewardModel
 
 
@@ -136,6 +134,8 @@ class RougeSimilarityFilter(FilterFunction):
     def __init__(
         self, existing_instructions: List[str], threshold: float = 0.7
     ):
+        from rouge import Rouge
+
         self.existing_instructions = existing_instructions
         self.threshold = threshold
         self.rouge = Rouge()

camel/datagen/source2synth/data_processor.py

@@ -15,8 +15,6 @@
 import random
 from typing import Any, Dict, List, Optional, Sequence
 
-from tqdm import tqdm
-
 from camel.agents.multi_hop_generator_agent import MultiHopGeneratorAgent
 from camel.datagen.source2synth.user_data_processor_config import (
     ProcessorConfig,
@@ -172,6 +170,8 @@ def construct_examples(
         logger.info("Starting to construct training examples...")
         examples = []
 
+        from tqdm import tqdm
+
         for data in tqdm(raw_data, desc="Constructing examples"):
             # 1. Text preprocessing
             processed_text = self._preprocess_text(data.get('text', ''))

camel/datasets/base_generator.py

@@ -20,7 +20,14 @@
 from typing import Any, Dict, List, Union
 
 from pydantic import ValidationError
-from torch.utils.data import IterableDataset
+
+try:
+    from torch.utils.data import IterableDataset
+except ImportError:
+
+    class IterableDataset:  # type: ignore[no-redef]
+        r"""Fallback when PyTorch is not installed."""
+        pass
 
 from camel.logger import get_logger
 

camel/datasets/static_dataset.py

@@ -16,6 +16,7 @@
 from collections.abc import Sequence
 from pathlib import Path
 from typing import (
+    TYPE_CHECKING,
     Any,
     Dict,
     List,
@@ -24,9 +25,20 @@
     Union,
 )
 
-from datasets import Dataset as HFDataset
 from pydantic import ValidationError
-from torch.utils.data import Dataset
+
+try:
+    from datasets import Dataset as HFDataset
+except ImportError:
+    HFDataset = None  # type: ignore[assignment,misc]
+
+try:
+    from torch.utils.data import Dataset
+except ImportError:
+
+    class Dataset:  # type: ignore[no-redef]
+        r"""Fallback when PyTorch is not installed."""
+        pass
 
 from camel.logger import get_logger
 
@@ -116,7 +128,7 @@ def _init_data(
             ValueError: If the Path has an unsupported file extension.
         """
 
-        if isinstance(data, HFDataset):
+        if HFDataset is not None and isinstance(data, HFDataset):
             raw_data = self._init_from_hf_dataset(data)
         elif isinstance(data, Dataset):
             raw_data = self._init_from_pytorch_dataset(data)