updates + format

This commit is contained in:
Awni Hannun 2023-12-14 12:09:10 -08:00
parent 29b7a97342
commit b1b9b11801
4 changed files with 55 additions and 56 deletions

View File

@ -1,9 +1,13 @@
# CIFAR and ResNets # CIFAR and ResNets
An example of training a ResNet on CIFAR-10 with MLX. Several ResNet configurations in accordance with the original [paper](https://arxiv.org/abs/1512.03385) are available. Also illustrates how to use `mlx-data` to download and load the dataset. An example of training a ResNet on CIFAR-10 with MLX. Several ResNet
configurations in accordance with the original
[paper](https://arxiv.org/abs/1512.03385) are available. The example also
illustrates how to use [MLX Data](https://github.com/ml-explore/mlx-data) to
load the dataset.
## Pre-requisites ## Pre-requisites
Install the dependencies: Install the dependencies:
``` ```
@ -11,6 +15,7 @@ pip install -r requirements.txt
``` ```
## Running the example ## Running the example
Run the example with: Run the example with:
``` ```
@ -29,23 +34,18 @@ For all available options, run:
python main.py --help python main.py --help
``` ```
## Throughput
On the tested device (M1 Macbook Pro, 16GB RAM), I get the following throughput with a `batch_size=256`:
```
Epoch: 0 | avg. tr_loss 2.074 | avg. tr_acc 0.216 | Train Throughput: 415.39 images/sec
```
When training on just the CPU (with the `--cpu` argument), the throughput is significantly lower (almost 30x!):
```
Epoch: 0 | avg. tr_loss 2.074 | avg. tr_acc 0.216 | Train Throughput: 13.5 images/sec
```
## Results ## Results
After training for 100 epochs, the following results were observed:
After training with the default `resnet20` architecture for 100 epochs, you
should see the following results:
``` ```
Epoch: 99 | avg. tr_loss 0.320 | avg. tr_acc 0.888 | Train Throughput: 416.77 images/sec Epoch: 99 | avg. Train loss 0.320 | avg. Train acc 0.888 | Throughput: 416.77 images/sec
Epoch: 99 | test_acc 0.807 Epoch: 99 | Test acc 0.807
``` ```
At the time of writing, `mlx` doesn't have in-built `schedulers`, nor a `BatchNorm` layer. We'll revisit this example for exact reproduction once these features are added.
Note this was run on an M1 Macbook Pro with 16GB RAM.
At the time of writing, `mlx` doesn't have built-in learning rate schedules,
nor a `BatchNorm` layer. We intend to update this example once these features
are added.

View File

@ -4,13 +4,15 @@ import math
def get_cifar10(batch_size, root=None): def get_cifar10(batch_size, root=None):
tr = load_cifar10(root=root) tr = load_cifar10(root=root)
num_tr_samples = tr.size()
mean = mx.array([0.485, 0.456, 0.406]).reshape((1, 1, 3)) mean = mx.array([0.485, 0.456, 0.406]).reshape((1, 1, 3))
std = mx.array([0.229, 0.224, 0.225]).reshape((1, 1, 3)) std = mx.array([0.229, 0.224, 0.225]).reshape((1, 1, 3))
def normalize(x):
x = x.astype("float32") / 255.0
return (x - mean) / std
tr_iter = ( tr_iter = (
tr.shuffle() tr.shuffle()
.to_stream() .to_stream()
@ -18,22 +20,11 @@ def get_cifar10(batch_size, root=None):
.pad("image", 0, 4, 4, 0.0) .pad("image", 0, 4, 4, 0.0)
.pad("image", 1, 4, 4, 0.0) .pad("image", 1, 4, 4, 0.0)
.image_random_crop("image", 32, 32) .image_random_crop("image", 32, 32)
.key_transform("image", lambda x: (x.astype("float32") / 255.0)) .key_transform("image", normalize)
.key_transform("image", lambda x: (x - mean) / std)
.batch(batch_size) .batch(batch_size)
) )
test = load_cifar10(root=root, train=False) test = load_cifar10(root=root, train=False)
num_test_samples = test.size() test_iter = test.to_stream().key_transform("image", normalize).batch(batch_size)
test_iter = (
test.to_stream()
.key_transform("image", lambda x: (x.astype("float32") / 255.0))
.key_transform("image", lambda x: (x - mean) / std)
.batch(batch_size)
)
num_tr_steps_per_epoch = num_tr_samples // batch_size
num_test_steps_per_epoch = num_test_samples // batch_size
return tr_iter, test_iter return tr_iter, test_iter

View File

@ -12,7 +12,8 @@ parser.add_argument(
"--arch", "--arch",
type=str, type=str,
default="resnet20", default="resnet20",
help="model architecture [resnet20, resnet32, resnet44, resnet56, resnet110, resnet1202]", choices=[f"resnet{d}" for d in [20, 32, 44, 56, 110, 1202]],
help="model architecture",
) )
parser.add_argument("--batch_size", type=int, default=256, help="batch size") parser.add_argument("--batch_size", type=int, default=256, help="batch size")
parser.add_argument("--epochs", type=int, default=100, help="number of epochs") parser.add_argument("--epochs", type=int, default=100, help="number of epochs")
@ -21,10 +22,6 @@ parser.add_argument("--seed", type=int, default=0, help="random seed")
parser.add_argument("--cpu", action="store_true", help="use cpu only") parser.add_argument("--cpu", action="store_true", help="use cpu only")
def loss_fn(model, inp, tgt):
return mx.mean(nn.losses.cross_entropy(model(inp), tgt))
def eval_fn(model, inp, tgt): def eval_fn(model, inp, tgt):
return mx.mean(mx.argmax(model(inp), axis=1) == tgt) return mx.mean(mx.argmax(model(inp), axis=1) == tgt)
@ -50,17 +47,25 @@ def train_epoch(model, train_iter, optimizer, epoch):
optimizer.update(model, grads) optimizer.update(model, grads)
mx.eval(model.parameters(), optimizer.state) mx.eval(model.parameters(), optimizer.state)
toc = time.perf_counter() toc = time.perf_counter()
loss_value = loss.item() loss = loss.item()
acc_value = acc.item() acc = acc.item()
losses.append(loss_value) losses.append(loss)
accs.append(acc_value) accs.append(acc)
samples_per_sec.append(x.shape[0] / (toc - tic)) throughput = x.shape[0] / (toc - tic)
samples_per_sec.append(throughput)
if batch_counter % 10 == 0: if batch_counter % 10 == 0:
print( print(
f"Epoch {epoch:02d} [{batch_counter:03d}] | tr_loss {loss_value:.3f} | tr_acc {acc_value:.3f} | Throughput: {x.shape[0] / (toc - tic):.2f} images/second" " | ".join(
(
f"Epoch {epoch:02d} [{batch_counter:03d}]",
f"Train loss {loss:.3f}",
f"Train acc {acc:.3f}",
f"Throughput: {throughput:.2f} images/second",
)
)
) )
mean_tr_loss = mx.mean(mx.array(losses)) eean_tr_loss = mx.mean(mx.array(losses))
mean_tr_acc = mx.mean(mx.array(accs)) mean_tr_acc = mx.mean(mx.array(accs))
samples_per_sec = mx.mean(mx.array(samples_per_sec)) samples_per_sec = mx.mean(mx.array(samples_per_sec))
return mean_tr_loss, mean_tr_acc, samples_per_sec return mean_tr_loss, mean_tr_acc, samples_per_sec
@ -81,24 +86,28 @@ def test_epoch(model, test_iter, epoch):
def main(args): def main(args):
mx.random.seed(args.seed) mx.random.seed(args.seed)
model = resnet.__dict__[args.arch]() model = getattr(resnet, args.arch)()
print("num_params: {:0.04f} M".format(model.num_params() / 1e6)) print("Number of params: {:0.04f} M".format(model.num_params() / 1e6))
mx.eval(model.parameters())
optimizer = optim.Adam(learning_rate=args.lr) optimizer = optim.Adam(learning_rate=args.lr)
train_data, test_data = get_cifar10(args.batch_size) train_data, test_data = get_cifar10(args.batch_size)
for epoch in range(args.epochs): for epoch in range(args.epochs):
epoch_tr_loss, epoch_tr_acc, train_throughput = train_epoch( tr_loss, tr_acc, throughput = train_epoch(model, train_data, optimizer, epoch)
model, train_data, optimizer, epoch
)
print( print(
f"Epoch: {epoch} | avg. tr_loss {epoch_tr_loss.item():.3f} | avg. tr_acc {epoch_tr_acc.item():.3f} | Train Throughput: {train_throughput.item():.2f} images/sec" " | ".join(
(
f"Epoch: {epoch}",
f"avg. Train loss {tr_loss.item():.3f}",
f"avg. Train acc {tr_acc.item():.3f}",
f"Throughput: {throughput.item():.2f} images/sec",
)
)
) )
epoch_test_acc = test_epoch(model, test_data, epoch) test_acc = test_epoch(model, test_data, epoch)
print(f"Epoch: {epoch} | test_acc {epoch_test_acc.item():.3f}") print(f"Epoch: {epoch} | Test acc {test_acc.item():.3f}")
train_data.reset() train_data.reset()
test_data.reset() test_data.reset()

View File

@ -59,7 +59,6 @@ class Block(nn.Module):
self.shortcut = None self.shortcut = None
def __call__(self, x): def __call__(self, x):
out = nn.relu(self.bn1(self.conv1(x))) out = nn.relu(self.bn1(self.conv1(x)))
out = self.bn2(self.conv2(out)) out = self.bn2(self.conv2(out))
if self.shortcut is None: if self.shortcut is None: