---

For something in between a pytorch and a karpathy/micrograd

This may not be the best deep learning framework, but it is a deep learning framework.

The Tensor class is a wrapper around a numpy array, except it does Tensor things.

Installation

pip3 install tinygrad --upgrade

Example

from tinygrad.tensor import Tensor

x = Tensor.eye(3)
y = Tensor([[2.0,0,-2.0]])
z = y.matmul(x).sum()
z.backward()

print(x.grad)  # dz/dx
print(y.grad)  # dz/dy

Same example in torch

import torch

x = torch.eye(3, requires_grad=True)
y = torch.tensor([[2.0,0,-2.0]], requires_grad=True)
z = y.matmul(x).sum()
z.backward()

print(x.grad)  # dz/dx
print(y.grad)  # dz/dy

Neural networks?

It turns out, a decent autograd tensor library is 90% of what you need for neural networks. Add an optimizer (SGD, RMSprop, and Adam implemented) from tinygrad.optim, write some boilerplate minibatching code, and you have all you need.

Neural network example (from test/test_mnist.py)

from tinygrad.tensor import Tensor
import tinygrad.optim as optim
from tinygrad.utils import layer_init_uniform

class TinyBobNet:
  def __init__(self):
    self.l1 = Tensor(layer_init_uniform(784, 128))
    self.l2 = Tensor(layer_init_uniform(128, 10))

  def forward(self, x):
    return x.dot(self.l1).relu().dot(self.l2).logsoftmax()

model = TinyBobNet()
optim = optim.SGD([model.l1, model.l2], lr=0.001)

# ... and complete like pytorch, with (x,y) data

out = model.forward(x)
loss = out.mul(y).mean()
loss.backward()
optim.step()

GPU Support?!

tinygrad supports GPUs through PyOpenCL. Not all ops are supported yet on the backward pass.

from tinygrad.tensor import Tensor
(Tensor.ones(4,4).cuda() + Tensor.ones(4,4).cuda()).cpu()

ImageNet inference

Despite being tiny, tinygrad supports the full EfficientNet. Pass in a picture to discover what it is.

python3 examples/efficientnet.py https://upload.wikimedia.org/wikipedia/commons/4/41/Chicken.jpg

protip: Set "GPU=1" environment variable if you want this to go faster.

The promise of small

tinygrad will always be below 1000 lines. If it isn't, we will revert commits until tinygrad becomes smaller.

Running tests

python -m pytest

TODO

• Train an EfficientNet on ImageNet
  • Make broadcasting work on the backward pass (simple please)
  • EfficientNet backward pass
  • Tensors on GPU (GPU support, must support Mac)
• Make tinygrad work on comma two and run driving model
• Reduce code
• Increase speed
• Add features