zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-06-24 17:31:16 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

gelu tanh approx (#989)

Browse Source 
* gelu tanh approx

* gelu tanh approx

* replace gelu approx with tanh approach

* fix comments

* fix comment
This commit is contained in:
Shiyu 2024-04-16 10:49:00 +08:00 committed by GitHub
parent cd9e184529
commit 107ba2891a
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 11 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
python/mlx/nn/layers/activations.py
View File

@ -163,15 +163,14 @@ def gelu_approx(x):
See :func:`gelu` for the exact computation. See :func:`gelu` for the exact computation.
This function approximates ``gelu`` with a maximum absolute error :math:`< This function approximates ``gelu`` with a maximum absolute error :math:`<
0.0003` in the range :math:`[-6, 6]` using the following 0.0005` in the range :math:`[-6, 6]` using the following
.. math:: .. math::
x = x \sigma\left(1.60033 x \left(1 + 0.0433603 x^2\right)\right) x = 0.5 * x * \left(1 + \text{Tanh}\left((\sqrt{2 / \pi} * \left(x + 0.044715 * x^3\right)\right)\right)
where :math:`\sigma(\cdot)` is the logistic sigmoid.
""" """
return x * mx.sigmoid(1.60033 * x * (1 + 0.0433603 * x.square())) return 0.5 * x * (1 + mx.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * x**3)))
@partial(mx.compile, shapeless=True) @partial(mx.compile, shapeless=True)
@ -504,7 +503,7 @@ class GELU(Module):
However, if ``approx`` is set to 'precise' or 'fast' it applies However, if ``approx`` is set to 'precise' or 'fast' it applies
.. math:: .. math::
\textrm{GELUApprox}(x) &= x * \sigma\left(1.60033 * x \left(1 + 0.0433603 * x^2\right)\right) \\ \textrm{GELUApprox}(x) &= 0.5 * x * \left(1 + \text{Tanh}\left((\sqrt{2 / \pi} * \left(x + 0.044715 * x^3\right)\right)\right) \\
\textrm{GELUFast}(x) &= x * \sigma\left(1.773 * x\right) \textrm{GELUFast}(x) &= x * \sigma\left(1.773 * x\right)
respectively. respectively.

8
python/tests/test_nn.py
View File

@ -717,16 +717,22 @@ class TestLayers(mlx_tests.MLXTestCase):
expected = np.array( expected = np.array(
[1.0093501, -0.16925684, 0.22918941, 0.60498625, 0.49459383] [1.0093501, -0.16925684, 0.22918941, 0.60498625, 0.49459383]
) )
# From: jax.nn.gelu(np.array(inputs), approximate=True)
expected_approx = np.array(
[1.0091482, -0.1693441, 0.22918446, 0.60491, 0.4945476]
)
out = nn.GELU()(mx.array(inputs)) out = nn.GELU()(mx.array(inputs))
self.assertTrue(np.allclose(out, expected)) self.assertTrue(np.allclose(out, expected))
out_approx = nn.GELU(approx="precise")(mx.array(inputs))
self.assertTrue(np.allclose(out_approx, expected_approx))
# Crudely check the approximations # Crudely check the approximations
x = mx.arange(-6.0, 6.0, 12 / 100) x = mx.arange(-6.0, 6.0, 12 / 100)
y = nn.gelu(x) y = nn.gelu(x)
y_hat1 = nn.gelu_approx(x) y_hat1 = nn.gelu_approx(x)
y_hat2 = nn.gelu_fast_approx(x) y_hat2 = nn.gelu_fast_approx(x)
self.assertLess(mx.abs(y - y_hat1).max(), 0.0003) self.assertLess(mx.abs(y - y_hat1).max(), 0.0005)
self.assertLess(mx.abs(y - y_hat2).max(), 0.025) self.assertLess(mx.abs(y - y_hat2).max(), 0.025)
def test_sin_pe(self): def test_sin_pe(self):