diff --git a/pyproject.toml b/pyproject.toml
index 7ccb0de0f62..e6b7f2755c8 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -209,4 +209,10 @@ filterwarnings = [
     # Tracking issue: https://github.com/huggingface/trl/issues/6435
     # Upstream fix (merged, unreleased as of liger-kernel 0.8.0): https://github.com/linkedin/Liger-Kernel/pull/1274
     "ignore:Error detected in IndexPutBackward0:UserWarning",
+
+    # bitsandbytes calls the deprecated torch._check_is_size in its CUDA quant ops (upstream, not actionable in TRL)
+    # Triggered indirectly by loading bitsandbytes-quantized models in the PEFT + quantization tests
+    # Tracking issue: https://github.com/huggingface/trl/issues/6447
+    # Upstream fix (merged, unreleased as of bitsandbytes 0.49.2): https://github.com/bitsandbytes-foundation/bitsandbytes/pull/1940
+    "ignore:_check_is_size will be removed in a future PyTorch release:FutureWarning",
 ]