zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-10-19 08:38:09 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Start to cleanup/unify accelerate and common back-ends (Part 1/N) (#1777)

Browse Source 
* start to cleanup/unify accelerate and common back-ends

* more progress

* simplify

* add half type and allow infs in simd exp

* unify softmax + quantized, more dispatches to simd quantized mm

* add sin/cos, use simd in vector-scalar ops

* faster CPU vectorize quant

* faster erf/erfinv
This commit is contained in:
Awni Hannun
2025-01-29 14:34:49 -08:00
committed by GitHub
parent 7064fed1b1
commit 4758c8baa1
47 changed files with 1920 additions and 2640 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/tests/test_quantized.py
View File

@@ -207,8 +207,8 @@ class TestQuantized(mlx_tests.MLXTestCase):
with self.subTest(shape=(B, M, N), group_size=group_size, bits=bits):
x_shape = (1, N) if B == 0 else (B, 1, N)
w_shape = (N, M) if B == 0 else (B, N, M)
x = mx.random.normal(shape=x_shape, key=k1)
w = mx.random.normal(shape=w_shape, key=k2)
x = 1e-1 * mx.random.normal(shape=x_shape, key=k1)
w = 1e-1 * mx.random.normal(shape=w_shape, key=k2)
w_q, scales, biases = mx.quantize(w, group_size, bits)
w_hat = mx.dequantize(w_q, scales, biases, group_size, bits)
y_q = mx.quantized_matmul(