George
George
Parse messy output from deepsparse.benchmark Also added commit sha info To run: `deepsparse.benchmark zoo:nlp/text_classification/bert-base/pytorch/huggingface/sst2/base-none -x benchmark.json` Before: https://github.com/neuralmagic/deepsparse/blob/main/src/deepsparse/benchmark/img/json_output.png After:
Given a git-server url (model page url, git clone http or git clone ssh), load its metadata ```python3 url = "https://git.neuralmagic.com/neuralmagic/cary" \ or " [email protected]:neuralmagic/cary.git" \ or "https://git.neuralmagic.com/neuralmagic/cary.git" server_model =...
Problem: Given a [model repo](https://git.neuralmagic.com/neuralmagic/cary), cannot get recipes*.md files from root ```python3 model = Model(model_path) files = model.files ``` Before: `files` does not contain recipes in root After: `files` contain...
* logger feature beanch * [V2 Logger] utils for import (#1536) * utils for import * clean up + tests * [V2 Logger] registry for default func/class (#1539) * registry...
server running pipeline for benchmark Note: Cannot have continuous batching with timer middleware. ## Configs ### Server side ```bash deepsparse.server --config_file config.yaml ``` ```yaml num_cores: 2 num_workers: 2 endpoints: -...
## Description Adds tests for Pipeline.run_async() ## Problem Testing run_async() currently requires some hacking in tests/server. Isolate Pipeline func's test. ## Solution Simple pipeline running run_async ## Usage ```python3 inference_state...
Show warning when overriding batch_size 0 to 1 https://app.asana.com/0/1201735099598270/1206262288703592
## Summary Bug fix for running custom dataset using accerate and use preprocessing func in module `src/sparseml/transformers/utils/preprocesing_funtions.py`, specified by string arg in parser. ```bash #!/bin/bash export CUDA_VISIBLE_DEVICES="0,1,2,3" NPROC=$(($(echo $CUDA_VISIBLE_DEVICES |...
Add coverage, will add it to nm-actions for github actions