-
[Integration of OpenTelemetry](https://opentelemetry-python-contrib.readthedocs.io/en/latest/instrumentation/grpc/grpc.html#usage-aio-server) with `tritonclient` can be made more convenient by passing…
-
**Is your feature request related to a problem? Please describe.**
No, it is not.
**Describe the solution you'd like**
Set `--backend-directory` to accept a list of 0, 1, or n. And 0 items for de…
-
**Description**
pb_utils.Tensor.from_dlpack throws a "DLPack tensor is not contiguous" exception on some (but not all) tensors that have a 0 dimension.
If any dimension is 0, I don't see how the t…
-
**Description**
We have a PyTorch model that we serve on the CPU using the triton inference server. We use the POLL approach to monitor the new model versions in the GCS bucket. We also use the gRPC …
-
Hello everyone,
I have 7 onnx models and run them with gpu_acceleration by tensorRT. But I cannot get the acceleration using concurrency loading models. Triton takes approximately 10 minutes to start…
-
The entrypoint of an `Application` class makes sense for the API, but the interface could be a lot more Pythonic. We should spike out a sensible `Application` class as the base of the non-COM API. Som…
-
I'd like to suggest some small set of icons that we could use to indicate what kinds of data files are contained within the various open data collections.
This would require a little bit of graphi…
-
A DLIS parameter may or may not have units. Unlike the `Channel` class, which has an attribute `units` that returns the channel's units as a string, the `Parameter` class has no attribute to return th…
-
When I read from S3 bucket which is actually a .dlis file:
`s3_client = boto3.client('s3')
s3_response_object = s3_client.get_object(Bucket=x['bucket'], Key=x['s3_key'])
object_content = s3_res…
-
### Describe the issue
Observing that repository checkout can consume a lot of time.
Due to model files examples stored in the repository.
Those files can be outdated to the project branch, but d…
mc-nv updated
4 months ago