Andrej micrograd Issues

karpathy/micrograd Issues

type annotation lacking/ maybe also add docstrings.
Updated 17 days ago
Topological sort - bug
Updated 20 days ago1
backward member implementation question
Updated a month ago
Adjusting parameters by sign and magnitude of gradient
Updated a month ago1
_backward as lambdas?
Updated a month ago3
Reseting the grad of weights and biases is not enough
Closed 4 months ago1
Regarding the gradient update of the __sub__ operation
Closed 5 months ago1
Ensure backward() is idempotent
Closed 7 months ago4
Another MiniGrad with the RAdam optimizer.
Updated 8 months ago
Vectorized modification with GPU support.
Updated 9 months ago
Vectorized implementation with PyTorch flavor
Updated a year ago
Noob question about backprop implementation
Closed 2 years ago4
Rename engine.py to value.py
Updated a year ago
A tensor version for micrograd inpired by this work
Updated a year ago
Zero_grad only zeros the weight and bias nodes, not the nodes for addition and multiplication
Closed a year ago1
Sequential MLP implementation
Updated a year ago
Issue with zero_grad?
Closed a year ago5
Homework Assignment Error with softmax activation function
Updated a year ago1
micrograd.NET: C# port for .NET developers
Updated 2 years ago
`other` should have a gradient in `__pow__` (?)
Closed 2 years ago2
Appreciation and seed to GPU Support using PyCuda
Updated 3 years ago
Grad should be a Value instead of python/numpy scalar
Updated 4 years ago2
PyPI package
Closed 4 years ago2
Would need 'substraction' support in Engine, mainly for regression loss functions like MSE
Closed 4 years ago4
Incorrect gradient when non-leaf Values are re-used
Closed 4 years ago3

karpathy / micrograd

karpathy/micrograd Issues

type annotation lacking/ maybe also add docstrings.

Topological sort - bug

backward member implementation question

Adjusting parameters by sign and magnitude of gradient

_backward as lambdas?

Reseting the grad of weights and biases is not enough

Regarding the gradient update of the sub operation

Ensure backward() is idempotent

Another MiniGrad with the RAdam optimizer.

Vectorized modification with GPU support.

Vectorized implementation with PyTorch flavor

Noob question about backprop implementation

Rename engine.py to value.py

A tensor version for micrograd inpired by this work

Zero_grad only zeros the weight and bias nodes, not the nodes for addition and multiplication

Sequential MLP implementation

Issue with zero_grad?

Homework Assignment Error with softmax activation function

micrograd.NET: C# port for .NET developers

`other` should have a gradient in `pow` (?)

Appreciation and seed to GPU Support using PyCuda

Grad should be a Value instead of python/numpy scalar

PyPI package

Would need 'substraction' support in Engine, mainly for regression loss functions like MSE

Incorrect gradient when non-leaf Values are re-used