Remove jit from readme and cleanup dependencies. (#1789)

Summary:
We do not support jit in the `test_bench.py` anymore. If anyone needs it, they will be suggested to implement their userbenchmarks for jit benchmarking.

We also want to simplify the user installation so that they don't need to install extra libraries except running the `install.py` script. Therefore, unnecessary package deps are removed from `install.py`.

Pull Request resolved: #1789

Reviewed By: msaroufim

Differential Revision: D47834999

Pulled By: xuzhao9

fbshipit-source-id: 7a069e4546679844265d34066b103308900c3a44

README.md

@@ -2,7 +2,7 @@
 This is a collection of open source benchmarks used to evaluate PyTorch performance.
 
 `torchbenchmark/models` contains copies of popular or exemplary workloads which have been modified to:
-*(a)* expose a standardized API for benchmark drivers, *(b)* optionally, enable JIT,
+*(a)* expose a standardized API for benchmark drivers, *(b)* optionally, enable backends such as torchinductor/torchscript,
  *(c)* contain a miniature version of train/test data and a dependency install script.
 
 ## Installation
@@ -36,11 +36,6 @@ Or use pip:
 pip install --pre torch torchvision torchtext torchaudio -i https://download.pytorch.org/whl/nightly/cu118
 ```
 
-Install other necessary libraries:
-```
-pip install boto3 pyyaml
-```
-
 Install the benchmark suite, which will recursively install dependencies for all the models.  Currently, the repo is intended to be installed from the source tree.
 ```
 git clone https://github.com/pytorch/benchmark
@@ -115,8 +110,8 @@ Some useful options include:
 - `--cpu_only` if running on a local CPU machine and ignoring machine configuration checks
 
 #### Examples of Benchmark Filters
-- `-k "test_train[NAME-cuda-jit]"` for a particular flavor of a particular model
-- `-k "(BERT and (not cuda) and (not jit))"` for a more flexible approach to filtering
+- `-k "test_train[NAME-cuda-eager]"` for a particular flavor of a particular model
+- `-k "(BERT and (not cuda))"` for a more flexible approach to filtering
 
 Note that `test_bench.py` will eventually be deprecated as the `userbenchmark` work evolve. Users are encouraged to explore and consider using [userbenchmark](#using-userbenchmark).
 
@@ -128,7 +123,7 @@ The `userbenchmark` allows you to develop your customized benchmarks with TorchB
 Sometimes you may want to just run train or eval on a particular model, e.g. for debugging or profiling.  Rather than relying on __main__ implementations inside each model, `run.py` provides a lightweight CLI for this purpose, building on top of the standard BenchmarkModel API.
 
 ```
-python run.py <model> [-d {cpu,cuda}] [-m {eager,jit}] [-t {eval,train}] [--profile]
+python run.py <model> [-d {cpu,cuda}] [-t {eval,train}] [--profile]
 ```
 Note: `<model>` can be a full, exact name, or a partial string match.
 

install.py

@@ -2,13 +2,10 @@
 import subprocess
 import os
 import sys
-import yaml
-import tarfile
 from utils import TORCH_DEPS, proxy_suggestion, get_pkg_versions, _test_https
 from pathlib import Path
 REPO_ROOT = Path(__file__).parent
 
-
 def pip_install_requirements(requirements_txt="requirements.txt"):
     if not _test_https():
         print(proxy_suggestion)