diff --git a/candle-pyo3/e5.py b/candle-pyo3/e5.py
index a0af0c5608..8ca48219de 100644
--- a/candle-pyo3/e5.py
+++ b/candle-pyo3/e5.py
@@ -60,7 +60,7 @@ def average_pool(last_hidden_states: torch.Tensor, attention_mask: torch.Tensor)
 
     loss = torch.nn.L1Loss()
     error = loss(hf_pooled, candle_pooled).mean().item()
-    print(f"Mean error between torch-referenze and candle: {error}")
+    print(f"Mean error between torch-reference and candle: {error}")
 
     # Quantize all attention 'weights'
     quantized_tensors = {}
@@ -101,4 +101,4 @@ def average_pool(last_hidden_states: torch.Tensor, attention_mask: torch.Tensor)
 
     candle_pooled_2 = average_pool(torch.tensor(encoder_out_2.values()), hf_tokenized["attention_mask"])
     error = loss(hf_pooled, candle_pooled_2).mean().item()
-    print(f"Mean error between torch-referenze and quantized-candle: {error}")
+    print(f"Mean error between torch-reference and quantized-candle: {error}")