various tests fixes for flakey tests (#2110)

* add mhenrichsen/alpaca_2k_test with revision dataset download fixture for flaky tests

* log slowest tests

* pin pynvml==11.5.3

* fix load local hub path

* optimize for speed w smaller models and val_set_size

* replace pynvml

* make the resume from checkpoint e2e faster

* make tests smaller

requirements.txt

@@ -26,7 +26,7 @@ numpy>=1.24.4,<=2.0.1
 evaluate==0.4.1
 scipy
 scikit-learn==1.4.2
-pynvml
+nvidia-ml-py==12.560.30
 art
 gradio==3.50.2
 tensorboard