feat: Support sending additional outputs from vLLM inference

README.md

@@ -203,6 +203,11 @@ you need to specify a different `shm-region-prefix-name` for each server. See
 [here](https://github.com/triton-inference-server/python_backend#running-multiple-instances-of-triton-server)
 for more information.
 
+## Additional vLLM outputs
+
+Additional vLLM outputs may be requested optionally on a per-request basis. See
+[this docs](docs/additional_outputs.md) for more information.
+
 ## Triton Metrics
 Starting with the 24.08 release of Triton, users can now obtain specific
 vLLM metrics by querying the Triton metrics endpoint (see complete vLLM metrics

ci/L0_additional_outputs_vllm/additional_outputs_test.py

@@ -0,0 +1,185 @@
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import json
+
+import numpy as np
+import pytest
+import tritonclient.grpc as grpcclient
+
+
+class TestAdditionalOutputs:
+    _grpc_url = "localhost:8001"
+    _model_name = "vllm_opt"
+    _sampling_parameters = {"temperature": "0", "top_p": "1"}
+    _prompt = "In this example,"
+
+    def _get_inputs(
+        self,
+        prompt,
+        stream=True,
+        sampling_parameters=None,
+        output_finish_reason=None,
+        output_cumulative_logprob=None,
+        output_num_token_ids=None,
+    ):
+        inputs = []
+
+        inputs.append(grpcclient.InferInput("text_input", [1], "BYTES"))
+        inputs[-1].set_data_from_numpy(
+            np.array([prompt.encode("utf-8")], dtype=np.object_)
+        )
+
+        inputs.append(grpcclient.InferInput("stream", [1], "BOOL"))
+        inputs[-1].set_data_from_numpy(np.array([stream], dtype=bool))
+
+        if sampling_parameters is not None:
+            inputs.append(grpcclient.InferInput("sampling_parameters", [1], "BYTES"))
+            inputs[-1].set_data_from_numpy(
+                np.array(
+                    [json.dumps(sampling_parameters).encode("utf-8")], dtype=np.object_
+                )
+            )
+
+        if output_finish_reason is not None:
+            inputs.append(grpcclient.InferInput("output_finish_reason", [1], "BOOL"))
+            inputs[-1].set_data_from_numpy(np.array([output_finish_reason], dtype=bool))
+
+        if output_cumulative_logprob is not None:
+            inputs.append(
+                grpcclient.InferInput("output_cumulative_logprob", [1], "BOOL")
+            )
+            inputs[-1].set_data_from_numpy(
+                np.array([output_cumulative_logprob], dtype=bool)
+            )
+
+        if output_num_token_ids is not None:
+            inputs.append(grpcclient.InferInput("output_num_token_ids", [1], "BOOL"))
+            inputs[-1].set_data_from_numpy(np.array([output_num_token_ids], dtype=bool))
+
+        return inputs
+
+    def _callback(self, result, error):
+        self._responses.append({"result": result, "error": error})
+
+    def _llm_infer(self, inputs):
+        self._responses = []
+        with grpcclient.InferenceServerClient(self._grpc_url) as client:
+            client.start_stream(self._callback)
+            client.async_stream_infer(
+                self._model_name, inputs=inputs, parameters=self._sampling_parameters
+            )
+            client.stop_stream()
+        assert len(self._responses) > 0
+
+    def _assert_text_output_valid(self):
+        text_output = ""
+        for response in self._responses:
+            result, error = response["result"], response["error"]
+            assert error is None
+            text_output += result.as_numpy(name="text_output")[0].decode("utf-8")
+        assert len(text_output) > 0, "output is empty"
+        assert text_output.count(" ") > 4, "output is not a sentence"
+
+    def _assert_finish_reason(self, output_finish_reason):
+        for i in range(len(self._responses)):
+            result, error = self._responses[i]["result"], self._responses[i]["error"]
+            assert error is None
+            finish_reason_np = result.as_numpy(name="finish_reason")
+            if output_finish_reason is None or output_finish_reason == False:
+                assert finish_reason_np is None
+                continue
+            finish_reason = finish_reason_np[0].decode("utf-8")
+            if i < len(self._responses) - 1:
+                assert finish_reason == "None"
+            else:
+                assert finish_reason == "length"
+
+    def _assert_cumulative_logprob(self, output_cumulative_logprob):
+        prev_cumulative_logprob = 0.0
+        for response in self._responses:
+            result, error = response["result"], response["error"]
+            assert error is None
+            cumulative_logprob_np = result.as_numpy(name="cumulative_logprob")
+            if output_cumulative_logprob is None or output_cumulative_logprob == False:
+                assert cumulative_logprob_np is None
+                continue
+            cumulative_logprob = cumulative_logprob_np[0].astype(float)
+            assert cumulative_logprob != prev_cumulative_logprob
+            prev_cumulative_logprob = cumulative_logprob
+
+    def _assert_num_token_ids(self, output_num_token_ids):
+        for response in self._responses:
+            result, error = response["result"], response["error"]
+            assert error is None
+            num_token_ids_np = result.as_numpy(name="num_token_ids")
+            if output_num_token_ids is None or output_num_token_ids == False:
+                assert num_token_ids_np is None
+                continue
+            num_token_ids = num_token_ids_np[0].astype(int)
+            # TODO: vLLM may return token ids identical to the previous one when
+            #       streaming, for example:
+            #
+            #       prev: None
+            #       curr: text=' the', token_ids=array('l', [5])
+            #
+            #       prev: text=' the', token_ids=array('l', [5, 1385])
+            #       curr: text=' the term', token_ids=array('l', [5, 1385])
+            #
+            #       prev: text=' the term', token_ids=array('l', [5, 1385, 44])
+            #       curr: text=' the term', token_ids=array('l', [5, 1385, 44])
+            #
+            #       prev: text=' the term', token_ids=array('l', [5, 1385, 44, 48])
+            #       curr: text=' the term “', token_ids=array('l', [5, 1385, 44, 48])
+            #
+            #       If this is no longer the case in a future release, change the assert
+            #       to assert num_token_ids > 0.
+            assert num_token_ids >= 0
+
+    @pytest.mark.parametrize("stream", [True, False])
+    @pytest.mark.parametrize("output_finish_reason", [None, True, False])
+    @pytest.mark.parametrize("output_cumulative_logprob", [None, True, False])
+    @pytest.mark.parametrize("output_num_token_ids", [None, True, False])
+    def test_additional_outputs(
+        self,
+        stream,
+        output_finish_reason,
+        output_cumulative_logprob,
+        output_num_token_ids,
+    ):
+        inputs = self._get_inputs(
+            self._prompt,
+            stream=stream,
+            sampling_parameters=self._sampling_parameters,
+            output_finish_reason=output_finish_reason,
+            output_cumulative_logprob=output_cumulative_logprob,
+            output_num_token_ids=output_num_token_ids,
+        )
+        self._llm_infer(inputs)
+        self._assert_text_output_valid()
+        self._assert_finish_reason(output_finish_reason)
+        self._assert_cumulative_logprob(output_cumulative_logprob)
+        self._assert_num_token_ids(output_num_token_ids)

ci/L0_additional_outputs_vllm/test.sh

@@ -0,0 +1,66 @@
+#!/bin/bash
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+export CUDA_VISIBLE_DEVICES=0
+source ../common/util.sh
+
+pip3 install pytest==8.1.1
+pip3 install tritonclient[grpc]
+
+# Prepare Model
+rm -rf models vllm_baseline_output.pkl && mkdir -p models
+SAMPLE_MODELS_REPO="../../samples/model_repository"
+cp -r $SAMPLE_MODELS_REPO/vllm_model models/vllm_opt
+sed -i 's/"gpu_memory_utilization": 0.5/"gpu_memory_utilization": 0.3/' models/vllm_opt/1/model.json
+
+RET=0
+
+# Test
+SERVER_LOG="vllm_opt.server.log"
+SERVER_ARGS="--model-repository=models"
+run_server
+if [ "$SERVER_PID" == "0" ]; then
+    echo -e "\n***\n*** Failed to start $SERVER\n***"
+    cat $SERVER_LOG
+    exit 1
+fi
+set +e
+python3 -m pytest --junitxml=test_additional_outputs.xml -s -v additional_outputs_test.py
+if [ $? -ne 0 ]; then
+    echo -e "\n***\n*** additional_outputs_test FAILED. \n***"
+    RET=1
+fi
+set -e
+kill $SERVER_PID
+wait $SERVER_PID
+
+if [ $RET -eq 0 ]; then
+    echo -e "\n***\n*** Test Passed\n***"
+else
+    echo -e "\n***\n*** Test FAILED\n***"
+fi
+exit $RET

ci/common/util.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright 2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright 2023-2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions
@@ -25,7 +25,7 @@
 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
-
+SERVER=${SERVER:=/opt/tritonserver/bin/tritonserver}
 SERVER_IPADDR=${TRITONSERVER_IPADDR:=localhost}
 SERVER_LOG=${SERVER_LOG:=./server.log}
 SERVER_TIMEOUT=${SERVER_TIMEOUT:=120}

docs/additional_outputs.md

@@ -0,0 +1,107 @@
+<!--
+# Copyright 2024, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+-->
+
+# Additional Outputs from vLLM
+
+The vLLM backend supports sending additional outputs from vLLM on top of the
+usual `text_output` when requested.
+
+All additional outputs are disabled by default and they need to be enabled on a
+per-request basis. If enabled, the corresponding output tensor will be set for
+all responses from the request.
+
+## Supported Additional Outputs
+
+### Finish Reason
+
+The reason why the sequence is finished. See
+[here](https://github.com/vllm-project/vllm/blob/v0.6.3.post1/vllm/outputs.py#L26)
+for more details.
+
+To enable, set `output_finish_reason` input tensor to `True`. The reason will be
+sent as a string on the `finish_reason` output tensor.
+
+Supported since r24.11.
+
+### Cumulative Log Probabilities
+
+The cumulative log probability of the generated output text. See
+[here](https://github.com/vllm-project/vllm/blob/v0.6.3.post1/vllm/outputs.py#L22)
+for more details.
+
+To enable, set `output_cumulative_logprob` input tensor to `True`. The floating
+point value will be sent on the `cumulative_logprob` output tensor.
+
+Supported since r24.11.
+
+### Number of token IDs
+
+The number of token IDs of the generated output text sent on this response. It
+is the difference in length of the token IDs generated from the last response to
+this response. If this is the first response, the last response length is
+presumed to be zero. See
+[here](https://github.com/vllm-project/vllm/blob/v0.6.3.post1/vllm/outputs.py#L21)
+for more details on the token IDs of the generated output text.
+
+To enable, set `output_num_token_ids` input tensor to `True`. The unsigned
+integer value will be sent on the `num_token_ids` output tensor.
+
+Supported since r24.11.
+
+## Examples
+
+### Add Finish Reason to Outputs
+
+```python
+import numpy as np
+import tritonclient.grpc as grpcclient
+
+inputs = []
+
+inputs.append(grpcclient.InferInput("text_input", [1], "BYTES"))
+inputs[-1].set_data_from_numpy(
+    np.array(["example prompt".encode("utf-8")], dtype=np.object_)
+)
+
+inputs.append(grpcclient.InferInput("output_finish_reason", [1], "BOOL"))
+inputs[-1].set_data_from_numpy(np.array([True], dtype=bool))
+
+def callback(result, error):
+    ...
+    print(result.as_numpy(name="finish_reason"))
+
+with grpcclient.InferenceServerClient("localhost:8001") as client:
+    client.start_stream(callback)
+    client.async_stream_infer("vLLM_model_name", inputs=inputs, ...)
+    client.stop_stream()
+```
+
+## Notes
+
+* Enabling additional outputs may impact performance, only add additional
+outputs when necessary.