github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-10 07:28:15 -05:00
Code Issues Packages Projects Releases Wiki Activity

docs: minor update to quickstart (#4746)

Browse Source 
update import, point to docs more, remove mention that assign to index not supported
This commit is contained in:
chenyu
2024-05-27 17:46:36 -04:00
committed by GitHub
parent 4cb38a15a5
commit db0e19fbd7
1 changed files with 5 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
docs/quickstart.md
View File

@@ -20,7 +20,7 @@ All high level operations in tinygrad operate on these tensors.
The tensor class can be imported like so:
```python
from tinygrad.tensor import Tensor
from tinygrad import Tensor
```
Tensors can be created from an existing data structure like a python list or numpy ndarray:
@@ -46,16 +46,16 @@ eye = Tensor.eye(3) # create a 3x3 identity matrix
arange = Tensor.arange(start=0, stop=10, step=1) # create a tensor of shape (10,) filled with values from 0 to 9
rand = Tensor.rand(2, 3) # create a tensor of shape (2, 3) filled with random values from a uniform distribution
randn = Tensor.randn(2, 3) # create a tensor of shape (2, 3) filled with random values from a normal distribution
randn = Tensor.randn(2, 3) # create a tensor of shape (2, 3) filled with random values from a standard normal distribution
uniform = Tensor.uniform(2, 3, low=0, high=10) # create a tensor of shape (2, 3) filled with random values from a uniform distribution between 0 and 10
```
There are even more of these factory methods, you can find them in the [Tensor](tensor.md) file.
All the tensors creation methods can take a `dtype` argument to specify the data type of the tensor.
All the tensors creation methods can take a `dtype` argument to specify the data type of the tensor, find the supported `dtype` in [dtypes](dtypes.md).
```python
from tinygrad.dtype import dtypes
from tinygrad import dtypes
t3 = Tensor([1, 2, 3, 4, 5], dtype=dtypes.int32)
```
@@ -139,7 +139,6 @@ def sparse_categorical_crossentropy(self, Y, ignore_index=-1) -> Tensor:
```
As we can see in this implementation of cross entropy loss, there are certain operations that tinygrad does not support natively.
Namely, operations that are load/store or assigning a value to a tensor at a certain index.
Load/store ops are not supported in tinygrad natively because they add complexity when trying to port to different backends, 90% of the models out there don't use/need them, and they can be implemented like it's done above with an `arange` mask.
For our optimizer we will be using the traditional stochastic gradient descent optimizer with a learning rate of 3e-4.
@@ -299,7 +298,7 @@ Many of the models in the [models/](https://github.com/tinygrad/tinygrad/tree/ma
There exist a bunch of environment variables that control the runtime behavior of tinygrad.
Some of the commons ones are `DEBUG` and the different backend enablement variables.
You can find a full list and their descriptions in [env_vars.md](https://github.com/tinygrad/tinygrad/blob/master/docs/env_vars.md).
You can find a full list and their descriptions in [env_vars.md](env_vars.md).
### Visualizing the Computation Graph