How to use the wandb.save function in wandb

def test_custom_file_policy_symlink(mocker, run_manager):
    mod = mocker.MagicMock()
    mocker.patch(
        'wandb.run_manager.FileEventHandlerThrottledOverwriteMinWait.on_modified', mod)
    with open("ckpt_0.txt", "w") as f:
        f.write("joy")
    with open("ckpt_1.txt", "w") as f:
        f.write("joy" * 100)
    wandb.save("ckpt_0.txt")
    with open("ckpt_0.txt", "w") as f:
        f.write("joy" * 100)
    wandb.save("ckpt_1.txt")
    run_manager.test_shutdown()
    assert isinstance(
        run_manager._file_event_handlers["ckpt_0.txt"], FileEventHandlerThrottledOverwriteMinWait)
    assert mod.called

def test_save_absolute_path(wandb_init_run):
    with open("/tmp/test.txt", "w") as f:
        f.write("something")
    wandb.save("/tmp/test.txt")
    assert os.path.exists(os.path.join(wandb_init_run.dir, "test.txt"))

args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

args.features_turned_on = sum([args.kle_stop, args.kle_rollback, args.gae, args.norm_obs, args.norm_returns, args.norm_adv, args.anneal_lr, args.clip_vloss, args.pol_layer_norm])
# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))

if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
# respect the default timelimit
assert isinstance(env.action_space, MultiDiscrete), "only MultiDiscrete action space is supported"
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):
    if int(args.episode_length):
        env._max_episode_steps = int(args.episode_length)
    args.episode_length = env._max_episode_steps
else:
    env = TimeLimit(env, int(args.episode_length))
env = NormalizedEnv(env.env, ob=args.norm_obs, ret=args.norm_returns, clipob=args.obs_clip, cliprew=args.rew_clip, gamma=args.gamma)
env = TimeLimit(env, int(args.episode_length))
random.seed(args.seed)

parser.add_argument('--ent-coef', type=float, default=0.01,
                       help="policy entropy's coefficient the loss function")
    args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))
if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
random.seed(args.seed)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
torch.backends.cudnn.deterministic = args.torch_deterministic
env.seed(args.seed)
env.action_space.seed(args.seed)
env.observation_space.seed(args.seed)
input_shape, preprocess_obs_fn = preprocess_obs_space(env.observation_space, device)
output_shape = preprocess_ac_space(env.action_space)
# respect the default timelimit
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):

args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

args.features_turned_on = sum([args.kle_stop, args.kle_rollback, args.gae, args.norm_obs, args.norm_returns, args.norm_adv, args.anneal_lr, args.clip_vloss, args.pol_layer_norm])
# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))

if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
# respect the default timelimit
assert isinstance(env.action_space, MultiDiscrete), "only MultiDiscrete action space is supported"
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):
    if int(args.episode_length):
        env._max_episode_steps = int(args.episode_length)
    args.episode_length = env._max_episode_steps
else:
    env = TimeLimit(env, int(args.episode_length))
env = NormalizedEnv(env.env, ob=args.norm_obs, ret=args.norm_returns, clipob=args.obs_clip, cliprew=args.rew_clip, gamma=args.gamma)
env = TimeLimit(env, int(args.episode_length))
random.seed(args.seed)

args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

args.features_turned_on = sum([args.kle_stop, args.kle_rollback, args.gae, args.norm_obs, args.norm_returns, args.norm_adv, args.anneal_lr, args.clip_vloss, args.pol_layer_norm])
# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))

if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
# respect the default timelimit
assert isinstance(env.action_space, MultiDiscrete), "only MultiDiscrete action space is supported"
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):
    if int(args.episode_length):
        env._max_episode_steps = int(args.episode_length)
    args.episode_length = env._max_episode_steps
else:
    env = TimeLimit(env, int(args.episode_length))
env = NormalizedEnv(env.env, ob=args.norm_obs, ret=args.norm_returns, clipob=args.obs_clip, cliprew=args.rew_clip, gamma=args.gamma)
env = TimeLimit(env, int(args.episode_length))
random.seed(args.seed)

args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

args.features_turned_on = sum([args.kle_stop, args.kle_rollback, args.gae, args.norm_obs, args.norm_returns, args.norm_adv, args.anneal_lr, args.clip_vloss, args.pol_layer_norm])
# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))

if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
# respect the default timelimit
assert isinstance(env.action_space, MultiDiscrete), "only MultiDiscrete action space is supported"
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):
    if int(args.episode_length):
        env._max_episode_steps = int(args.episode_length)
    args.episode_length = env._max_episode_steps
else:
    env = TimeLimit(env, int(args.episode_length))
env = NormalizedEnv(env.env, ob=args.norm_obs, ret=args.norm_returns, clipob=args.obs_clip, cliprew=args.rew_clip, gamma=args.gamma)
env = TimeLimit(env, int(args.episode_length))
random.seed(args.seed)

args = parser.parse_args()
    if not args.seed:
        args.seed = int(time.time())

args.features_turned_on = sum([args.kle_stop, args.kle_rollback, args.gae, args.norm_obs, args.norm_returns, args.norm_adv, args.anneal_lr, args.clip_vloss, args.pol_layer_norm])
# TRY NOT TO MODIFY: setup the environment
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))

if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, entity=args.wandb_entity, tensorboard=True, config=vars(args), name=experiment_name, monitor_gym=True)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

# TRY NOT TO MODIFY: seeding
device = torch.device('cuda' if torch.cuda.is_available() and args.cuda else 'cpu')
env = gym.make(args.gym_id)
# respect the default timelimit
assert isinstance(env.action_space, MultiDiscrete), "only MultiDiscrete action space is supported"
assert isinstance(env, TimeLimit) or int(args.episode_length), "the gym env does not have a built in TimeLimit, please specify by using --episode-length"
if isinstance(env, TimeLimit):
    if int(args.episode_length):
        env._max_episode_steps = int(args.episode_length)
    args.episode_length = env._max_episode_steps
else:
    env = TimeLimit(env, int(args.episode_length))
env = NormalizedEnv(env.env, ob=args.norm_obs, ret=args.norm_returns, clipob=args.obs_clip, cliprew=args.rew_clip, gamma=args.gamma)
env = TimeLimit(env, int(args.episode_length))
random.seed(args.seed)

for index, vector in tqdm(enumerate(code_representations)):
            if vector is not None:
                indices.add_item(index, vector)
        indices.build(10)

        for query in queries:
            for idx, _ in zip(*query_model(query, model, indices, language)):
                predictions.append((query, language, definitions[idx]['identifier'], definitions[idx]['url']))

    df = pd.DataFrame(predictions, columns=['query', 'language', 'identifier', 'url'])
    df.to_csv(predictions_csv, index=False)

    if run_id:
        # upload model predictions CSV file to W&B
        wandb.init(id=run_id, resume="must")
        wandb.save(predictions_csv)

returns.append( ret)
        lengths.append( t)

    return returns, lengths

# TRY NOT TO MODIFY: start the game
experiment_name = f"{args.gym_id}__{args.exp_name}__{args.seed}__{int(time.time())}"
writer = SummaryWriter(f"runs/{experiment_name}")
writer.add_text('hyperparameters', "|param|value|\n|-|-|\n%s" % (
        '\n'.join([f"|{key}|{value}|" for key, value in vars(args).items()])))
if args.prod_mode:
    import wandb
    wandb.init(project=args.wandb_project_name, tensorboard=True, config=vars(args), name=experiment_name)
    writer = SummaryWriter(f"/tmp/{experiment_name}")
    wandb.save(os.path.abspath(__file__))

global_step = 0
while global_step < args.total_timesteps:
    next_obs = np.array(env.reset())

    # MODIFIED: Keeping track of train episode returns and lengths
    train_episode_return = 0.
    train_episode_length = 0

    done = False

    # TRY NOT TO MODIFY: prepare the execution of the game.
    for step in range(args.episode_length):
        global_step += 1
        obs = next_obs.copy()