Auto Image Classification

First, install gradsflow

pip install git+https://github.com/gradsflow/gradsflow@main

In [ ]:

import os
import sys
import warnings
from pathlib import Path

warnings.filterwarnings("ignore")

os.chdir("../../")

import os import sys import warnings from pathlib import Path warnings.filterwarnings("ignore") os.chdir("../../")

In [ ]:

import ray
from flash.core.data.utils import download_data
from gradsflow import AutoImageClassifier
from gradsflow.data.image import image_dataset_from_directory

import ray from flash.core.data.utils import download_data from gradsflow import AutoImageClassifier from gradsflow.data.image import image_dataset_from_directory

Let's use Hymenoptera dataset provided by Flash which contain images of Ants and Bees for creating Image Classication Model.

In [ ]:

data_dir = "/Users/aniket/personal/gradsflow/gradsflow/data/"  # replace with your filepath
# download_data("https://pl-flash-data.s3.amazonaws.com/hymenoptera_data.zip", data_dir)

data_dir = "/Users/aniket/personal/gradsflow/gradsflow/data/" # replace with your filepath # download_data("https://pl-flash-data.s3.amazonaws.com/hymenoptera_data.zip", data_dir)

In [ ]:

train_data = image_dataset_from_directory(f"{data_dir}/hymenoptera_data/train/", transform=True)
train_dl = train_data.dataloader

val_data = image_dataset_from_directory(f"{data_dir}/hymenoptera_data/val/", transform=True)
val_dl = val_data.dataloader

train_data = image_dataset_from_directory(f"{data_dir}/hymenoptera_data/train/", transform=True) train_dl = train_data.dataloader val_data = image_dataset_from_directory(f"{data_dir}/hymenoptera_data/val/", transform=True) val_dl = val_data.dataloader

If you want to run Gradsflow on a remote server then first setup ray cluster and initialize ray with the remote address.

In [ ]:

# ray.init(address="REMOTE_IP_ADDR")
# ray.init(local_mode=True)

# ray.init(address="REMOTE_IP_ADDR") # ray.init(local_mode=True)

To train an image classifier create an object of AutoImageClassifier and provide number of trials and timeout.

In [ ]:

model = AutoImageClassifier(
    train_dataloader=train_dl,
    val_dataloader=val_dl,
    num_classes=2,
    n_trials=1,
    optimization_metric="train_loss",
    timeout=50,
)
print("AutoImageClassifier initialised!")

model = AutoImageClassifier( train_dataloader=train_dl, val_dataloader=val_dl, num_classes=2, n_trials=1, optimization_metric="train_loss", timeout=50, ) print("AutoImageClassifier initialised!")

AutoImageClassifier initialised!

In [ ]:

analysis = model.hp_tune()
print("completed!")

analysis = model.hp_tune() print("completed!")

2021-10-08 11:48:37,675	INFO services.py:1263 -- View the Ray dashboard at http://127.0.0.1:8265
2021-10-08 11:48:40,110	WARNING function_runner.py:558 -- Function checkpointing is disabled. This may result in unexpected behavior when using checkpointing features or certain schedulers. To enable, set the train function arguments to be `func(config, checkpoint_dir=None)`.

== Status ==
Memory usage on this node: 10.4/16.0 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/8 CPUs, 0/0 GPUs, 0.0/4.54 GiB heap, 0.0/2.27 GiB objects
Result logdir: /Users/aniket/ray_results/optimization_objective_2021-10-08_11-48-40
Number of trials: 1/1 (1 PENDING)

Trial name	status	loc	backbone	lr	optimizer
optimization_objective_94628_00000	PENDING		ssl_resnet50	0.00387828	sgd

2021-10-08 11:52:31,654	WARNING tune.py:518 -- SIGINT received (e.g. via Ctrl+C), ending Ray Tune run. This will try to checkpoint the experiment state one last time. Press CTRL+C one more time (or send SIGINT/SIGKILL/SIGTERM) to skip. 
2021-10-08 11:52:31,665	ERROR trial_runner.py:773 -- Trial optimization_objective_94628_00000: Error processing event.
Traceback (most recent call last):
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/tune/trial_runner.py", line 739, in _process_trial
    results = self.trial_executor.fetch_result(trial)
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/tune/ray_trial_executor.py", line 746, in fetch_result
    result = ray.get(trial_future[0], timeout=DEFAULT_GET_TIMEOUT)
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/_private/client_mode_hook.py", line 82, in wrapper
    return func(*args, **kwargs)
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/worker.py", line 1621, in get
    raise value.as_instanceof_cause()
ray.exceptions.RayTaskError: ray::ImplicitFunc.train_buffered() (pid=80657, ip=192.168.40.84, repr=<ray.tune.function_runner.ImplicitFunc object at 0x7fbb0274f4f0>)
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/tune/trainable.py", line 178, in train_buffered
    result = self.train()
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/tune/trainable.py", line 237, in train
    result = self.step()
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/site-packages/ray/tune/function_runner.py", line 361, in step
    result = self._results_queue.get(
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/queue.py", line 179, in get
    self.not_empty.wait(remaining)
  File "/Users/aniket/miniconda3/envs/am/lib/python3.8/threading.py", line 306, in wait
    gotit = waiter.acquire(True, timeout)
KeyboardInterrupt

During handling of the above exception, another exception occurred:

ray::ImplicitFunc.train_buffered() (pid=80657, ip=192.168.40.84, repr=<ray.tune.function_runner.ImplicitFunc object at 0x7fbb0274f4f0>)
ray.exceptions.TaskCancelledError: Task: TaskID(69a6825d641b46131cef336a2abb8d3ef89d181901000000) was cancelled
2021-10-08 11:52:31,685	INFO stopper.py:348 -- Reached timeout of 50 seconds. Stopping all trials.

Result for optimization_objective_94628_00000:
  {}
  

== Status ==
Memory usage on this node: 9.3/16.0 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/8 CPUs, 0/0 GPUs, 0.0/4.54 GiB heap, 0.0/2.27 GiB objects
Result logdir: /Users/aniket/ray_results/optimization_objective_2021-10-08_11-48-40
Number of trials: 1/1 (1 ERROR)

Trial name	status	loc	backbone	lr	optimizer
optimization_objective_94628_00000	ERROR		ssl_resnet50	0.00387828	sgd

Number of errored trials: 1

Trial name	# failures	error file
optimization_objective_94628_00000	1	/Users/aniket/ray_results/optimization_objective_2021-10-08_11-48-40/optimization_objective_94628_00000_0_backbone=ssl_resnet50,lr=0.0038783,optimizer=sgd_2021-10-08_11-48-40/error.txt

== Status ==
Memory usage on this node: 9.2/16.0 GiB
Using FIFO scheduling algorithm.
Resources requested: 0/8 CPUs, 0/0 GPUs, 0.0/4.54 GiB heap, 0.0/2.27 GiB objects
Result logdir: /Users/aniket/ray_results/optimization_objective_2021-10-08_11-48-40
Number of trials: 1/1 (1 ERROR)

Trial name	status	loc	backbone	lr	optimizer
optimization_objective_94628_00000	ERROR		ssl_resnet50	0.00387828	sgd

Number of errored trials: 1

Trial name	# failures	error file
optimization_objective_94628_00000	1	/Users/aniket/ray_results/optimization_objective_2021-10-08_11-48-40/optimization_objective_94628_00000_0_backbone=ssl_resnet50,lr=0.0038783,optimizer=sgd_2021-10-08_11-48-40/error.txt

2021-10-08 11:52:31,819	ERROR tune.py:557 -- Trials did not complete: [optimization_objective_94628_00000]
2021-10-08 11:52:31,822	INFO tune.py:561 -- Total run time: 231.72 seconds (231.52 seconds for the tuning loop).
2021-10-08 11:52:31,824	WARNING tune.py:565 -- Experiment has been interrupted, but the most recent state was saved. You can continue running this experiment by passing `resume=True` to `tune.run()`
2021-10-08 11:52:31,837	WARNING experiment_analysis.py:644 -- Could not find best trial. Did you pass the correct `metric` parameter?

completed!

In [ ]:

# ray.shutdown()

# ray.shutdown()

(pid=80657) /Users/aniket/miniconda3/envs/am/lib/python3.8/multiprocessing/resource_tracker.py:216: UserWarning: resource_tracker: There appear to be 32 leaked semaphore objects to clean up at shutdown
(pid=80657)   warnings.warn('resource_tracker: There appear to be %d '

---

Last update: October 8, 2021