diff --git a/slime/utils/data.py b/slime/utils/data.py
index 7a0f5b67bf..43c47f2b0f 100644
--- a/slime/utils/data.py
+++ b/slime/utils/data.py
@@ -1,6 +1,6 @@
 import random
 
-from datasets import Dataset as hf_ds
+from pandas as pd
 
 from slime.utils.types import Sample
 
@@ -9,15 +9,14 @@ __all__ = ["Dataset"]
 
 # TODO: don't read the whole file into memory.
 def read_file(path):
-    if path.endswith(".jsonl") or path.endswith(".json"):
-        ds = hf_ds.from_json(path)
+    if path.endswith(".jsonl"):
+        df = pd.read_json(path, lines=True)
     elif path.endswith(".parquet"):
-        ds = hf_ds.from_parquet(path)
+        df = pd.read_parquet(path, dtype_backend="pyarrow")
     else:
         raise ValueError(f"Unsupported file format: {path}. Supported formats are .jsonl and .parquet.")
-
-    for data in ds:
-        yield data
+    for _, row in df.iterrows():
+        yield row.to_dict()
 
 
 class Dataset:
@@ -40,6 +39,11 @@ class Dataset:
             if apply_chat_template:
                 if tool_key is not None:
                     tools = data[tool_key]
+                    if isinstance(tools, str):
+                        tools = json.loads(tools)
+                    elif isinstance(tools, np.ndarray):
+                        tools =  tools.tolist()
+                    assert isinstance(tools, list), f"tools must be a list, got {type(tools)} instead"
                 else:
                     tools = None
                 prompt = tokenizer.apply_chat_template(prompt, tools, tokenize=False, add_generation_prompt=True)