This is an automated email from the ASF dual-hosted git repository.

jackietien pushed a commit to branch force_ci/object_type
in repository https://gitbox.apache.org/repos/asf/iotdb.git

commit 8b3a3d42a8971704c2437614e40d5105289b624c
Merge: efc633100da b4c542dc1e1
Author: JackieTien97 <[email protected]>
AuthorDate: Thu Dec 18 11:52:25 2025 +0800

    resolve conflicts with master

 iotdb-core/ainode/iotdb/ainode/core/inference/inference_request_pool.py | 1 -
 iotdb-core/ainode/iotdb/ainode/core/model/sundial/pipeline_sundial.py   | 2 +-
 .../queryengine/plan/relational/function/tvf/ForecastTableFunction.java | 2 ++
 3 files changed, 3 insertions(+), 2 deletions(-)

diff --cc 
iotdb-core/ainode/iotdb/ainode/core/inference/inference_request_pool.py
index 3cca9b183c8,c31bcd3d762..0bc1ab76bed
--- a/iotdb-core/ainode/iotdb/ainode/core/inference/inference_request_pool.py
+++ b/iotdb-core/ainode/iotdb/ainode/core/inference/inference_request_pool.py
@@@ -141,10 -140,7 +141,9 @@@ class InferenceRequestPool(mp.Process)
                      # more infer kwargs can be added here
                  )
              else:
 +                batch_output = None
                  self._logger.error("[Inference] Unsupported pipeline type.")
 +            batch_output = self._inference_pipeline.postprocess(batch_output)
- 
              offset = 0
              for request in requests:
                  request.output_tensor = request.output_tensor.to(self.device)
diff --cc iotdb-core/ainode/iotdb/ainode/core/model/sundial/pipeline_sundial.py
index 69422dfadb2,ee128802d24..2c786cbbc81
--- a/iotdb-core/ainode/iotdb/ainode/core/model/sundial/pipeline_sundial.py
+++ b/iotdb-core/ainode/iotdb/ainode/core/model/sundial/pipeline_sundial.py
@@@ -50,12 -45,7 +50,12 @@@ class SundialPipeline(ForecastPipeline)
              num_samples=num_samples,
              revin=revin,
          )
 -        return self._postprocess(output)
 -
 -    def _postprocess(self, output: torch.Tensor):
 -        return output.mean(dim=1)
 +        return outputs
 +
 +    def postprocess(self, outputs: torch.Tensor):
 +        """
 +        The outputs shape should be 3D, we need to take the mean value across 
num_samples dimension and expand dims.
 +        """
 +        outputs = outputs.mean(dim=1).unsqueeze(1)
 +        outputs = super().postprocess(outputs)
-         return outputs
++        return outputs

Reply via email to