Skip to content

Commit 2f046b8

Browse files
mc-nvmc-nv
authored
Update README and add RELEASE notes for 23.03 (#5556)
* Update README and add RELEASE notes for 23.03 * Addressed comments * Remove statment related to 'main' branch * Added dot
1 parent b86fa83 commit 2f046b8

File tree

2 files changed

+327
-3
lines changed

2 files changed

+327
-3
lines changed

‎README.md

+224-3
Original file line numberDiff line numberDiff line change
@@ -28,6 +28,227 @@
2828

2929
# Triton Inference Server
3030

31-
> **Note** <br>
32-
> You are currently on the r23.03 branch which tracks stabilization
33-
> towards the next release.<br> This branch is not usable during stabilization.
31+
[![License](https://img.shields.io/badge/License-BSD3-lightgrey.svg)](https://opensource.org/licenses/BSD-3-Clause)
32+
33+
----
34+
Triton Inference Server is an open source inference serving software that
35+
streamlines AI inferencing. Triton enables teams to deploy any AI model from
36+
multiple deep learning and machine learning frameworks, including TensorRT,
37+
TensorFlow, PyTorch, ONNX, OpenVINO, Python, RAPIDS FIL, and more. Triton
38+
supports inference across cloud, data center,edge and embedded devices on NVIDIA
39+
GPUs, x86 and ARM CPU, or AWS Inferentia. Triton delivers optimized performance
40+
for many query types, including real time, batched, ensembles and audio/video
41+
streaming.
42+
43+
Major features include:
44+
45+
- [Supports multiple deep learning
46+
frameworks](https://github.com/triton-inference-server/backend#where-can-i-find-all-the-backends-that-are-available-for-triton)
47+
- [Supports multiple machine learning
48+
frameworks](https://github.com/triton-inference-server/fil_backend)
49+
- [Concurrent model
50+
execution](docs/user_guide/architecture.md#concurrent-model-execution)
51+
- [Dynamic batching](docs/user_guide/model_configuration.md#dynamic-batcher)
52+
- [Sequence batching](docs/user_guide/model_configuration.md#sequence-batcher) and
53+
[implicit state management](docs/user_guide/architecture.md#implicit-state-management)
54+
for stateful models
55+
- Provides [Backend API](https://github.com/triton-inference-server/backend) that
56+
allows adding custom backends and pre/post processing operations
57+
- Model pipelines using
58+
[Ensembling](docs/user_guide/architecture.md#ensemble-models) or [Business
59+
Logic Scripting
60+
(BLS)](https://github.com/triton-inference-server/python_backend#business-logic-scripting)
61+
- [HTTP/REST and GRPC inference
62+
protocols](docs/customization_guide/inference_protocols.md) based on the community
63+
developed [KServe
64+
protocol](https://github.com/kserve/kserve/tree/master/docs/predict-api/v2)
65+
- A [C API](docs/customization_guide/inference_protocols.md#in-process-triton-server-api) and
66+
[Java API](docs/customization_guide/inference_protocols.md#java-bindings-for-in-process-triton-server-api)
67+
allow Triton to link directly into your application for edge and other in-process use cases
68+
- [Metrics](docs/user_guide/metrics.md) indicating GPU utilization, server
69+
throughput, server latency, and more
70+
71+
**New to Triton Inference Server?** Make use of
72+
[these tutorials](https://github.com/triton-inference-server/tutorials)
73+
to begin your Triton journey!
74+
75+
Join the [Triton and TensorRT community](https://www.nvidia.com/en-us/deep-learning-ai/triton-tensorrt-newsletter/) and
76+
stay current on the latest product updates, bug fixes, content, best practices,
77+
and more. Need enterprise support? NVIDIA global support is available for Triton
78+
Inference Server with the
79+
[NVIDIA AI Enterprise software suite](https://www.nvidia.com/en-us/data-center/products/ai-enterprise/).
80+
81+
## Serve a Model in 3 Easy Steps
82+
83+
```bash
84+
# Step 1: Create the example model repository
85+
git clone -b r23.03 https://github.com/triton-inference-server/server.git
86+
cd server/docs/examples
87+
./fetch_models.sh
88+
89+
# Step 2: Launch triton from the NGC Triton container
90+
docker run --gpus=1 --rm --net=host -v ${PWD}/model_repository:/models nvcr.io/nvidia/tritonserver:23.03-py3 tritonserver --model-repository=/models
91+
92+
# Step 3: Sending an Inference Request
93+
# In a separate console, launch the image_client example from the NGC Triton SDK container
94+
docker run -it --rm --net=host nvcr.io/nvidia/tritonserver:23.03-py3-sdk
95+
/workspace/install/bin/image_client -m densenet_onnx -c 3 -s INCEPTION /workspace/images/mug.jpg
96+
97+
# Inference should return the following
98+
Image '/workspace/images/mug.jpg':
99+
15.346230 (504) = COFFEE MUG
100+
13.224326 (968) = CUP
101+
10.422965 (505) = COFFEEPOT
102+
```
103+
Please read the [QuickStart](docs/getting_started/quickstart.md) guide for additional information
104+
regarding this example. The quickstart guide also contains an example of how to launch Triton on [CPU-only systems](docs/getting_started/quickstart.md#run-on-cpu-only-system). New to Triton and wondering where to get started? Watch the [Getting Started video](https://youtu.be/NQDtfSi5QF4).
105+
106+
## Examples and Tutorials
107+
108+
Check out [NVIDIA LaunchPad](https://www.nvidia.com/en-us/data-center/products/ai-enterprise-suite/trial/)
109+
for free access to a set of hands-on labs with Triton Inference Server hosted on
110+
NVIDIA infrastructure.
111+
112+
Specific end-to-end examples for popular models, such as ResNet, BERT, and DLRM
113+
are located in the
114+
[NVIDIA Deep Learning Examples](https://github.com/NVIDIA/DeepLearningExamples)
115+
page on GitHub. The
116+
[NVIDIA Developer Zone](https://developer.nvidia.com/nvidia-triton-inference-server)
117+
contains additional documentation, presentations, and examples.
118+
119+
## Documentation
120+
121+
### Build and Deploy
122+
123+
The recommended way to build and use Triton Inference Server is with Docker
124+
images.
125+
126+
- [Install Triton Inference Server with Docker containers](docs/customization_guide/build.md#building-with-docker) (*Recommended*)
127+
- [Install Triton Inference Server without Docker containers](docs/customization_guide/build.md#building-without-docker)
128+
- [Build a custom Triton Inference Server Docker container](docs/customization_guide/compose.md)
129+
- [Build Triton Inference Server from source](docs/customization_guide/build.md#building-on-unsupported-platforms)
130+
- [Build Triton Inference Server for Windows 10](docs/customization_guide/build.md#building-for-windows-10)
131+
- Examples for deploying Triton Inference Server with Kubernetes and Helm on [GCP](deploy/gcp/README.md),
132+
[AWS](deploy/aws/README.md), and [NVIDIA FleetCommand](deploy/fleetcommand/README.md)
133+
134+
### Using Triton
135+
136+
#### Preparing Models for Triton Inference Server
137+
138+
The first step in using Triton to serve your models is to place one or
139+
more models into a [model repository](docs/user_guide/model_repository.md). Depending on
140+
the type of the model and on what Triton capabilities you want to enable for
141+
the model, you may need to create a [model
142+
configuration](docs/user_guide/model_configuration.md) for the model.
143+
144+
- [Add custom operations to Triton if needed by your model](docs/user_guide/custom_operations.md)
145+
- Enable model pipelining with [Model Ensemble](docs/user_guide/architecture.md#ensemble-models)
146+
and [Business Logic Scripting (BLS)](https://github.com/triton-inference-server/python_backend#business-logic-scripting)
147+
- Optimize your models setting [scheduling and batching](docs/user_guide/architecture.md#models-and-schedulers)
148+
parameters and [model instances](docs/user_guide/model_configuration.md#instance-groups).
149+
- Use the [Model Analyzer tool](https://github.com/triton-inference-server/model_analyzer)
150+
to help optimize your model configuration with profiling
151+
- Learn how to [explicitly manage what models are available by loading and
152+
unloading models](docs/user_guide/model_management.md)
153+
154+
#### Configure and Use Triton Inference Server
155+
156+
- Read the [Quick Start Guide](docs/getting_started/quickstart.md) to run Triton Inference
157+
Server on both GPU and CPU
158+
- Triton supports multiple execution engines, called
159+
[backends](https://github.com/triton-inference-server/backend#where-can-i-find-all-the-backends-that-are-available-for-triton), including
160+
[TensorRT](https://github.com/triton-inference-server/tensorrt_backend),
161+
[TensorFlow](https://github.com/triton-inference-server/tensorflow_backend),
162+
[PyTorch](https://github.com/triton-inference-server/pytorch_backend),
163+
[ONNX](https://github.com/triton-inference-server/onnxruntime_backend),
164+
[OpenVINO](https://github.com/triton-inference-server/openvino_backend),
165+
[Python](https://github.com/triton-inference-server/python_backend), and more
166+
- Not all the above backends are supported on every platform supported by Triton.
167+
Look at the
168+
[Backend-Platform Support Matrix](https://github.com/triton-inference-server/backend/blob/main/docs/backend_platform_support_matrix.md)
169+
to learn which backends are supported on your target platform.
170+
- Learn how to [optimize performance](docs/user_guide/optimization.md) using the
171+
[Performance Analyzer](https://github.com/triton-inference-server/client/blob/main/src/c++/perf_analyzer/README.md)
172+
and
173+
[Model Analyzer](https://github.com/triton-inference-server/model_analyzer)
174+
- Learn how to [manage loading and unloading models](docs/user_guide/model_management.md) in
175+
Triton
176+
- Send requests directly to Triton with the [HTTP/REST JSON-based
177+
or gRPC protocols](docs/customization_guide/inference_protocols.md#httprest-and-grpc-protocols)
178+
179+
#### Client Support and Examples
180+
181+
A Triton *client* application sends inference and other requests to Triton. The
182+
[Python and C++ client libraries](https://github.com/triton-inference-server/client)
183+
provide APIs to simplify this communication.
184+
185+
- Review client examples for [C++](https://github.com/triton-inference-server/client/blob/main/src/c%2B%2B/examples),
186+
[Python](https://github.com/triton-inference-server/client/blob/main/src/python/examples),
187+
and [Java](https://github.com/triton-inference-server/client/blob/main/src/java/src/main/java/triton/client/examples)
188+
- Configure [HTTP](https://github.com/triton-inference-server/client#http-options)
189+
and [gRPC](https://github.com/triton-inference-server/client#grpc-options)
190+
client options
191+
- Send input data (e.g. a jpeg image) directly to Triton in the [body of an HTTP
192+
request without any additional metadata](https://github.com/triton-inference-server/server/blob/main/docs/protocol/extension_binary_data.md#raw-binary-request)
193+
194+
### Extend Triton
195+
196+
[Triton Inference Server's architecture](docs/user_guide/architecture.md) is specifically
197+
designed for modularity and flexibility
198+
199+
- [Customize Triton Inference Server container](docs/customization_guide/compose.md) for your use case
200+
- [Create custom backends](https://github.com/triton-inference-server/backend)
201+
in either [C/C++](https://github.com/triton-inference-server/backend/blob/main/README.md#triton-backend-api)
202+
or [Python](https://github.com/triton-inference-server/python_backend)
203+
- Create [decouple backends and models](docs/user_guide/decoupled_models.md) that can send
204+
multiple responses for a request or not send any responses for a request
205+
- Use a [Triton repository agent](docs/customization_guide/repository_agents.md) to add functionality
206+
that operates when a model is loaded and unloaded, such as authentication,
207+
decryption, or conversion
208+
- Deploy Triton on [Jetson and JetPack](docs/user_guide/jetson.md)
209+
- [Use Triton on AWS
210+
Inferentia](https://github.com/triton-inference-server/python_backend/tree/main/inferentia)
211+
212+
### Additional Documentation
213+
214+
- [FAQ](docs/user_guide/faq.md)
215+
- [User Guide](docs/README.md#user-guide)
216+
- [Customization Guide](docs/README.md#customization-guide)
217+
- [Release Notes](https://docs.nvidia.com/deeplearning/triton-inference-server/release-notes/index.html)
218+
- [GPU, Driver, and CUDA Support
219+
Matrix](https://docs.nvidia.com/deeplearning/dgx/support-matrix/index.html)
220+
221+
## Contributing
222+
223+
Contributions to Triton Inference Server are more than welcome. To
224+
contribute please review the [contribution
225+
guidelines](CONTRIBUTING.md). If you have a backend, client,
226+
example or similar contribution that is not modifying the core of
227+
Triton, then you should file a PR in the [contrib
228+
repo](https://github.com/triton-inference-server/contrib).
229+
230+
## Reporting problems, asking questions
231+
232+
We appreciate any feedback, questions or bug reporting regarding this project.
233+
When posting [issues in GitHub](https://github.com/triton-inference-server/server/issues),
234+
follow the process outlined in the [Stack Overflow document](https://stackoverflow.com/help/mcve).
235+
Ensure posted examples are:
236+
- minimal – use as little code as possible that still produces the
237+
same problem
238+
- complete – provide all parts needed to reproduce the problem. Check
239+
if you can strip external dependencies and still show the problem. The
240+
less time we spend on reproducing problems the more time we have to
241+
fix it
242+
- verifiable – test the code you're about to provide to make sure it
243+
reproduces the problem. Remove all other problems that are not
244+
related to your request/question.
245+
246+
For issues, please use the provided bug report and feature request templates.
247+
248+
For questions, we recommend posting in our community
249+
[GitHub Discussions.](https://github.com/triton-inference-server/server/discussions)
250+
251+
## For more information
252+
253+
Please refer to the [NVIDIA Developer Triton page](https://developer.nvidia.com/nvidia-triton-inference-server)
254+
for more information.

‎RELEASE.md

+103
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,103 @@
1+
<!--
2+
# Copyright 2018-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3+
#
4+
# Redistribution and use in source and binary forms, with or without
5+
# modification, are permitted provided that the following conditions
6+
# are met:
7+
# * Redistributions of source code must retain the above copyright
8+
# notice, this list of conditions and the following disclaimer.
9+
# * Redistributions in binary form must reproduce the above copyright
10+
# notice, this list of conditions and the following disclaimer in the
11+
# documentation and/or other materials provided with the distribution.
12+
# * Neither the name of NVIDIA CORPORATION nor the names of its
13+
# contributors may be used to endorse or promote products derived
14+
# from this software without specific prior written permission.
15+
#
16+
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
17+
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18+
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
19+
# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
20+
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
21+
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
22+
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
23+
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
24+
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25+
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26+
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27+
-->
28+
29+
# Release Notes for 2.32.0
30+
31+
## New Freatures and Improvements
32+
33+
* Added the
34+
[Parameters Extension](https://github.com/triton-inference-server/server/blob/main/docs/protocol/extension_parameters.md)
35+
which allows an inference request to provide custom parameters that cannot be
36+
provided as inputs. These parameters can be used in the python backend as
37+
described
38+
[here](https://github.com/triton-inference-server/python_backend#inference-request-parameters).
39+
40+
* Added support for models that use decoupled API for Business Scripting Logic
41+
(BLS) in Python backend. Examples can be found
42+
[here](https://github.com/triton-inference-server/python_backend/blob/main/examples/decoupled/README.md).
43+
44+
* The same model name can be used across different repositories if the
45+
`--model-namespacing` flag is set.
46+
47+
* Triton’s Response Cache feature has been converted internally to a shared
48+
library implementation of the new
49+
[TRITONCACHE APIs](https://github.com/triton-inference-server/core/blob/main/include/triton/core/tritoncache.h),
50+
similar to how backends and repo agents are used today. The default cache
51+
implementation is
52+
[local_cache](https://github.com/triton-inference-server/local_cache), which
53+
is equivalent to the fixed-size in-memory buffer implementation used before.
54+
The `--response-cache-byte-size` flag will continue to function in the same
55+
way, but the `--cache-config` flag will be the preferred method of cache
56+
configuration moving forward. For more information, see the cache
57+
documentation
58+
[here](https://github.com/triton-inference-server/server/blob/main/docs/user_guide/response_cache.md).
59+
60+
* Triton’s
61+
[trace tool](https://github.com/triton-inference-server/server/blob/main/docs/user_guide/trace.md)
62+
now supports tracing for `request_id`.
63+
64+
* Refer to the 23.03 column of the
65+
[Frameworks Support Matrix](https://docs.nvidia.com/deeplearning/frameworks/support-matrix/index.html)
66+
for container image versions on which the 23.03 inference server container is
67+
based.
68+
69+
## Known Issues
70+
71+
* Support for TensorFlow1 will be removed starting from 23.04.
72+
73+
* Some systems which implement `malloc()` may not release memory back to the
74+
operating system right away causing a false memory leak. This can be mitigated
75+
by using a different malloc implementation. Tcmalloc is installed in the
76+
Triton container and can be
77+
[used by specifying the library in LD_PRELOAD](https://github.com/triton-inference-server/server/blob/r23.03/docs/user_guide/model_management.md#model-control-mode-explicit).
78+
79+
* Auto-complete may cause an increase in server start time. To avoid a start
80+
time increase, users can provide the full model configuration and launch the
81+
server with `--disable-auto-complete-config`.
82+
83+
* Auto-complete does not support PyTorch models due to lack of metadata in the
84+
model. It can only verify that the number of inputs and the input names
85+
matches what is specified in the model configuration. There is no model
86+
metadata about the number of outputs and datatypes. Related PyTorch bug:
87+
https://github.com/pytorch/pytorch/issues/38273
88+
89+
* Triton Client PIP wheels for ARM SBSA are not available from PyPI and pip will
90+
install an incorrect Jetson version of Triton Client library for Arm SBSA.
91+
92+
The correct client wheel file can be pulled directly from the Arm SBSA SDK
93+
image and manually installed.
94+
95+
* Traced models in PyTorch seem to create overflows when int8 tensor values are
96+
transformed to `int32` on the GPU.
97+
98+
Refer to https://github.com/pytorch/pytorch/issues/66930 for more information.
99+
100+
* Triton cannot retrieve GPU metrics with MIG-enabled GPU devices (A100 and A30).
101+
102+
* Triton metrics might not work if the host machine is running a separate DCGM
103+
agent on bare-metal or in a container.

0 commit comments

Comments
 (0)