mrbean
mrbean
### Feature request I think it would be awesome to be able to easily train a Tesla style HydraNet but using a transformer backbone. The model would take a model_id...
It would be great to be able to ingest data into a Weaviate with a Spark DataFrameWriter. This would allow easier integration for those who are currently using Lakehouse architectures...
### Feature request Add the ability to turn a model into a [FasterTransformer](https://github.com/NVIDIA/FasterTransformer) and use that to run inference. ### Motivation This is faster than ONNX in a lot of...
### System Info ```shell Building from source Running inside the nvcr.io/nvidia/tensorrt:21.07-py3 docker container ``` (note there is a bug so you will have to build from #286) ### Who can...
# What does this PR do? Allows to disable shape inference in quantization. This is the fix suggested in [this issue](https://github.com/microsoft/onnxruntime/issues/11072) but there was currently no way to leverage this...
### System Info ```shell Databricks runtime running on GPU Spark cluster Built optimum from source Python 3.8 ``` ### Who can help? @JingyaHuang @philschmid ### Information - [ ] The...
### System Info ```shell Built from source Building on Databricks ``` ### Who can help? @lewtun @michaelbenayoun ### Information - [X] The official example scripts - [X] My own modified...
# What does this PR do? Allow users to pass in their own provider to the `from_pretrained` method and spike it through to the load_model call in `ORTModel`. Currently the...
# What does this PR do? Fix the activation type for quantizing for tensorrt Fixes ```bash ValueError: Invalid combination of use_static_quantization = False and activations_dtype = QuantType.QInt8. OnnxRuntime dynamic quantization...
I would like to create a custom pipeline as seen [here](https://huggingface.co/docs/transformers/v4.19.2/en/main_classes/pipelines#pipeline-custom-code) that can leverage a quantized model. I don't see the capability to extend some sort of `OrtPipeline` class in...