systems
systems copied to clipboard
Merlin Systems provides tools for combining recommendation models with other elements of production recommender systems (like feature stores, nearest neighbor search, and exploration strategies) into...
This PR will introduce the initial hugectr Operator. This operator works along and will need a wrapper operator to handle inputs coming from a dataframe. The PR lays the foundation...
This PR is the first in a series of refactors. It isolates all tests involving triton into one sub folder in the unit tests directory.
This PR moves the triton tests from the ops test folder into the runtimes/triton test folder. This will make it easier to move all triton tests, because they will be...
# 🚀 Feature request It'd be nice to avoid all these DeprecationWarnings when we run `from merlin.systems.dag.ensemble import Ensemble`. ``` /usr/local/lib/python3.8/dist-packages/tritonclient/grpc/model_config_pb2.py:19: DeprecationWarning: Call to deprecated create function FileDescriptor(). Note: Create...
### Bug description ### Steps/Code to reproduce bug I run [this example from Merlin Models](https://github.com/NVIDIA-Merlin/models/blob/main/examples/usecases/transformers-next-item-prediction.ipynb). I run it all the way (and inclusive) to the cell where I train the...
- [ ] Use DLpack to leverage GPU memory between triton models in ensemble - [ ] Upgrade numpy in containers and see if dlpack works with Triton tensors -...
- [ ] Create a numpy/cupy dispatch mechanism (like pandas/cudf in NVT) - [ ] Apply DLpack to pass GPU tensors from Python back-end to other models - [ ]...