python_backend pytorch example as_numpy() error

[flian2]

Description

I’m using triton python_backend to run the pytorch example in the python_backend repo. I packaged the pytroch dependencies into a conda environment and can load the model successfully. However when running the client inference script provided in the repo, I encounter the following error when trying to get the output from the httpclient response. It seems that the response is empty.

---

ValueError                                Traceback (most recent call last)
Cell In[20], line 33
     30 response = client.infer(model_name, inputs, request_id=str(1), outputs=outputs)
     32 result = response.get_response()
---> 33 output0_data = response.as_numpy("OUTPUT0")
     34 output1_data = response.as_numpy("OUTPUT1")
     36 print(
     37     "INPUT0 ({}) + INPUT1 ({}) = OUTPUT0 ({})".format(
     38         input0_data, input1_data, output0_data
     39     )
     40 )

File ~/SageMaker/custom-miniconda/miniconda/envs/custom_python310/lib/python3.10/site-packages/tritonclient/http/_infer_result.py:208, in InferResult.as_numpy(self, name)
    204             if not has_binary_data:
    205                 np_array = np.array(
    206                     output["data"], dtype=triton_to_np_dtype(datatype)
    207                 )
--> 208             np_array = np_array.reshape(output["shape"])
    209             return np_array
    210 return None

ValueError: cannot reshape array of size 0 into shape (4,)

I tried running the add_sub example and the response.as_numpy("OUTPUT0") worked fine with the expected output.

Triton Information What version of Triton are you using?

server_version 2.41.0

Are you using the Triton container or did you build it yourself?

I’m using a sagemaker docker image for triton server: 763104351884.dkr.ecr.us-east-1.amazonaws.com/sagemaker-tritonserver:23.12-py3

To Reproduce

The model.py

import json

# triton_python_backend_utils is available in every Triton Python model. You
# need to use this module to create inference requests and responses. It also
# contains some utility functions for extracting information from model_config
# and converting Triton input/output types to numpy types.
import triton_python_backend_utils as pb_utils
from torch import nn

class AddSubNet(nn.Module):
    """
    Simple AddSub network in PyTorch. This network outputs the sum and
    subtraction of the inputs.
    """

    def __init__(self):
        super(AddSubNet, self).__init__()

    def forward(self, input0, input1):
        return (input0 + input1), (input0 - input1)

class TritonPythonModel:
    """Your Python model must use the same class name. Every Python model
    that is created must have "TritonPythonModel" as the class name.
    """

    def initialize(self, args):
        """`initialize` is called only once when the model is being loaded.
        Implementing `initialize` function is optional. This function allows
        the model to initialize any state associated with this model.

        Parameters
        ----------
        args : dict
          Both keys and values are strings. The dictionary keys and values are:
          * model_config: A JSON string containing the model configuration
          * model_instance_kind: A string containing model instance kind
          * model_instance_device_id: A string containing model instance device ID
          * model_repository: Model repository path
          * model_version: Model version
          * model_name: Model name
        """

        # You must parse model_config. JSON string is not parsed here
        self.model_config = model_config = json.loads(args["model_config"])

        # Get OUTPUT0 configuration
        output0_config = pb_utils.get_output_config_by_name(model_config, "OUTPUT0")

        # Get OUTPUT1 configuration
        output1_config = pb_utils.get_output_config_by_name(model_config, "OUTPUT1")

        # Convert Triton types to numpy types
        self.output0_dtype = pb_utils.triton_string_to_numpy(
            output0_config["data_type"]
        )
        self.output1_dtype = pb_utils.triton_string_to_numpy(
            output1_config["data_type"]
        )

        # Instantiate the PyTorch model
        self.add_sub_model = AddSubNet()

    def execute(self, requests):
        """`execute` must be implemented in every Python model. `execute`
        function receives a list of pb_utils.InferenceRequest as the only
        argument. This function is called when an inference is requested
        for this model. Depending on the batching configuration (e.g. Dynamic
        Batching) used, `requests` may contain multiple requests. Every
        Python model, must create one pb_utils.InferenceResponse for every
        pb_utils.InferenceRequest in `requests`. If there is an error, you can
        set the error argument when creating a pb_utils.InferenceResponse.

        Parameters
        ----------
        requests : list
          A list of pb_utils.InferenceRequest

        Returns
        -------
        list
          A list of pb_utils.InferenceResponse. The length of this list must
          be the same as `requests`
        """

        output0_dtype = self.output0_dtype
        output1_dtype = self.output1_dtype

        responses = []

        # Every Python backend must iterate over everyone of the requests
        # and create a pb_utils.InferenceResponse for each of them.
        for request in requests:
            # Get INPUT0
            in_0 = pb_utils.get_input_tensor_by_name(request, "INPUT0")
            # Get INPUT1
            in_1 = pb_utils.get_input_tensor_by_name(request, "INPUT1")

            out_0, out_1 = self.add_sub_model(in_0.as_numpy(), in_1.as_numpy())

            # Create output tensors. You need pb_utils.Tensor
            # objects to create pb_utils.InferenceResponse.
            out_tensor_0 = pb_utils.Tensor("OUTPUT0", out_0.astype(output0_dtype))
            out_tensor_1 = pb_utils.Tensor("OUTPUT1", out_1.astype(output1_dtype))

            # Create InferenceResponse. You can set an error here in case
            # there was a problem with handling this inference request.
            # Below is an example of how you can set errors in inference
            # response:
            #
            # pb_utils.InferenceResponse(
            #    output_tensors=..., TritonError("An error occurred"))
            inference_response = pb_utils.InferenceResponse(
                output_tensors=[out_tensor_0, out_tensor_1]
            )
            responses.append(inference_response)

        # You should return a list of pb_utils.InferenceResponse. Length
        # of this list must match the length of `requests` list.
        return responses

    def finalize(self):
        """`finalize` is called only once when the model is being unloaded.
        Implementing `finalize` function is optional. This function allows
        the model to perform any necessary clean ups before exit.
        """
        print("Cleaning up...")

config.pbtxt:

name: "add_sub"
backend: "python"

input [
  {
    name: "INPUT0"
    data_type: TYPE_FP32
    dims: [ 4 ]
  }
]
input [
  {
    name: "INPUT1"
    data_type: TYPE_FP32
    dims: [ 4 ]
  }
]
output [
  {
    name: "OUTPUT0"
    data_type: TYPE_FP32
    dims: [ 4 ]
  }
]
output [
  {
    name: "OUTPUT1"
    data_type: TYPE_FP32
    dims: [ 4 ]
  }
]

instance_group [{ kind: KIND_CPU }]

The client.py:

import sys

import numpy as np
import tritonclient.http as httpclient
from tritonclient.utils import *

model_name = "add_sub"
shape = [4]

with httpclient.InferenceServerClient("localhost:8000") as client:
    input0_data = np.random.rand(*shape).astype(np.float32)
    input1_data = np.random.rand(*shape).astype(np.float32)
    inputs = [
        httpclient.InferInput(
            "INPUT0", input0_data.shape, np_to_triton_dtype(input0_data.dtype)
        ),
        httpclient.InferInput(
            "INPUT1", input1_data.shape, np_to_triton_dtype(input1_data.dtype)
        ),
    ]

    inputs[0].set_data_from_numpy(input0_data)
    inputs[1].set_data_from_numpy(input1_data)

    outputs = [
        httpclient.InferRequestedOutput("OUTPUT0"),
        httpclient.InferRequestedOutput("OUTPUT1"),
    ]

    response = client.infer(model_name, inputs, request_id=str(1), outputs=outputs)

    result = response.get_response()
    output0_data = response.as_numpy("OUTPUT0")
    output1_data = response.as_numpy("OUTPUT1")

    print(
        "INPUT0 ({}) + INPUT1 ({}) = OUTPUT0 ({})".format(
            input0_data, input1_data, output0_data
        )
    )
    print(
        "INPUT0 ({}) - INPUT1 ({}) = OUTPUT1 ({})".format(
            input0_data, input1_data, output1_data
        )
    )

    if not np.allclose(input0_data + input1_data, output0_data):
        print("add_sub example error: incorrect sum")
        sys.exit(1)

    if not np.allclose(input0_data - input1_data, output1_data):
        print("add_sub example error: incorrect difference")
        sys.exit(1)

    print("PASS: add_sub")
    sys.exit(0)

Expected behavior I expect that the client script

output0_data = response.as_numpy("OUTPUT0")
output1_data = response.as_numpy("OUTPUT1")

will convert the output tensors in the http response to numpy array.

[oandreeva-nv]

Hi @flian2 , I've tried the reproducer in 24.08 and everything went through. Could you please check 24.08 as well?

[flian2]

Hi @oandreeva-nv , thanks! I'll give it a try.

[flian2]

Hi @oandreeva-nv, I tried this image nvcr.io/nvidia/tritonserver:24.08-py3 but I see the same error when running the client code. Could you tell me your tritonclient version when you reproduced? Thanks!

[mauriciocm9]

Any updates? I'm having the same issue with these versions: 24.08-py3, 24.09-py3 24.02-py3. Maybe is it related with aws? I'm also running my docker container inside an ec2 with the amazon linux ami.

[mauriciocm9]

Ok, apparently i was able to find the issue. Maybe your conda pack is installing a version of numpy that is conflicting with torch or a different package. In my case i was packing a numpy version higher than 2, which it isn't supported by torch.

[KrishnanPrash]

Hey @mauriciocm9 @flian2, Hopefully, this PR should resolve the issue you are experiencing allowing you to install numpy 1.x or numpy 2.x in your environment without any issues from the python backend. This fix will tentatively be a part of the 24.11 release.

For now, as @mauriciocm9 recommended, I would restrict the version of numpy in your conda environment to: numpy<2

[KrishnanPrash]

Closing this issue due to inactivity. Please feel free to reopen if needed.