
*Please install TVM by fetching the following commit id from the master branch                                                                                             * Commit ID:  5a27632e274fff57087ed0b6eb2856b6e5946cfb

* Please follow the instuctions in ./libxsmm_wrapper/README to install the libxsmm implementation and wrapper
* Run the script 'mb1_tuned_latest.py' as follows
  * $LD_PRELOAD=./libxsmm_wrapper/libxsmm_wrapper.so python -u mb1_tuned_latest.py -d <layer_name> 
  * layer_name can be any layer from resnet2,resnet3, ..., resnet20
  * These layers are from resnet-50 with minibatch size =1
  * Eg.
    $LD_PRELOAD=./libxsmm_wrapper/libxsmm_wrapper.so python -u mb1_tuned_latest.py -d resnet3 
* A sample slurm job script 'resnet3.slurm' is given to run on a cluster. Please run using:
    $sbatch resnet3.slurm
* Tuning can be paused, by commenting out the tuning function in the script and just measuring the best parameters
* contained in log file. The place to comment out is line no. 406 in mb1_tuned_latest.py
* Result containing the best performance is recorded in a generated excel sheet. 
