@@ -45,7 +45,7 @@ any C++ code.
4545 - [ ` initialize ` ] ( #initialize )
4646 - [ ` execute ` ] ( #execute )
4747 - [ Default Mode] ( #default-mode )
48- - [ Error Handling] ( #error-handling )
48+ - [ Error Handling] ( #error-handling )
4949 - [ Decoupled mode] ( #decoupled-mode )
5050 - [ Use Cases] ( #use-cases )
5151 - [ Known Issues] ( #known-issues )
@@ -62,8 +62,8 @@ any C++ code.
6262 - [ Running Multiple Instances of Triton Server] ( #running-multiple-instances-of-triton-server )
6363- [ Business Logic Scripting] ( #business-logic-scripting )
6464 - [ Using BLS with Decoupled Models] ( #using-bls-with-decoupled-models )
65- - [ Using BLS with Stateful Models] ( #using-bls-with-stateful-models )
6665 - [ Model Loading API] ( #model-loading-api )
66+ - [ Using BLS with Stateful Models] ( #using-bls-with-stateful-models )
6767 - [ Limitation] ( #limitation )
6868- [ Interoperability and GPU Support] ( #interoperability-and-gpu-support )
6969 - [ ` pb_utils.Tensor.to_dlpack() -> PyCapsule ` ] ( #pb_utilstensorto_dlpack---pycapsule )
@@ -72,7 +72,9 @@ any C++ code.
7272 - [ Input Tensor Device Placement] ( #input-tensor-device-placement )
7373- [ Frameworks] ( #frameworks )
7474 - [ PyTorch] ( #pytorch )
75+ - [ PyTorch Determinism] ( #pytorch-determinism )
7576 - [ TensorFlow] ( #tensorflow )
77+ - [ TensorFlow Determinism] ( #tensorflow-determinism )
7678- [ Custom Metrics] ( #custom-metrics )
7779- [ Examples] ( #examples )
7880 - [ AddSub in NumPy] ( #addsub-in-numpy )
@@ -81,7 +83,8 @@ any C++ code.
8183 - [ Business Logic Scripting] ( #business-logic-scripting-1 )
8284 - [ Preprocessing] ( #preprocessing )
8385 - [ Decoupled Models] ( #decoupled-models )
84- - [ Auto-complete Config] ( #auto-complete-config )
86+ - [ Model Instance Kind] ( #model-instance-kind )
87+ - [ Auto-complete config] ( #auto-complete-config )
8588 - [ Custom Metrics] ( #custom-metrics-1 )
8689- [ Running with Inferentia] ( #running-with-inferentia )
8790- [ Logging] ( #logging )
@@ -678,7 +681,7 @@ above.
678681If you want to create a tar file that contains all your Python dependencies or
679682you want to use different Python environments for each Python model you need to
680683create a * Custom Execution Environment* in Python backend.
681- Currently, Python backend only supports
684+ Currently, Python backend supports
682685[ conda-pack] ( https://conda.github.io/conda-pack/ ) for this purpose.
683686[ conda-pack] ( https://conda.github.io/conda-pack/ ) ensures that your conda
684687environment is portable. You can create a tar file for your conda environment
@@ -704,7 +707,17 @@ If this variable is not exported and similar packages are installed outside your
704707conda environment, your tar file may not contain all the dependencies required
705708for an isolated Python environment.
706709
707- After creating the tar file from the conda environment, you need to tell Python
710+ Alternatively, Python backend also supports unpacked conda execution
711+ environments, given it points to an activation script to setup the conda
712+ environment. To do this, the execution environment can be first packed using
713+ [ conda-pack] ( https://conda.github.io/conda-pack/ ) and then unpacked, or created
714+ using [ conda create -p] ( https://docs.conda.io/projects/conda/en/latest/commands/create.html ) .
715+ In this case, the conda activation script is located in:
716+ ``` $path_to_conda_pack/lib/python<your.python.version>/site-packages/conda_pack/scripts/posix/activate ```
717+ This speeds up the server loading time for models.
718+
719+ After creating the packed file from the conda environment or creating a conda
720+ environment with a custom activation script, you need to tell Python
708721backend to use that environment for your model. You can do this by adding the
709722lines below to the ` config.pbtxt ` file:
710723
0 commit comments