Fix fp_qmv_impl small-output-dim branch using raw fp8 scale byte

mlx/backend/metal/kernels/fp_quantized.h

@@ -440,7 +440,7 @@ METAL_FUNC void fp_qmv_impl(
         auto wl = (const device uint8_t*)(ws + row * in_vec_size_w);
         const device auto* sl = scales + row * in_vec_size_g;
 
-        uint8_t s = sl[0];
+        U s = dequantize_scale<U, group_size>(sl[0]);
         result[row] += qdot<U, values_per_thread, bits>(wl, x_thread, s);
       }
 

python/tests/test_quantized.py

@@ -482,6 +482,47 @@ def test_fp_qmv(self):
             self.assertEqual(y_q.shape, y_hat.shape)
             self.assertLess((y_q - y_hat).abs().max(), 1e-3)
 
+    def test_fp_qmv_small_non_multiples(self):
+        # Regression test for the fp_qmv_impl out_vec_size < 8 branch, whose
+        # full-block loop loaded the fp8 scale as a raw byte instead of decoding
+        # it through dequantize_scale (see fp_quantized.h). The bug only bites
+        # fp-quantized matvec when the output dim is < 8 AND K is large enough to
+        # run at least one full block (K > block_size), so it slipped past both
+        # test_fp_qmv (output dim >= 8) and test_qmv_small_non_multiples (K = 32,
+        # below block_size, exercising only the correct remainder loop).
+        # K = 512 forces full blocks; N in {1..7} hits the < 8 branch.
+        key = mx.random.key(0)
+        k1, k2 = mx.random.split(key)
+        K = 512
+        for mode, group_size, bits in [
+            ("mxfp4", 32, 4),
+            ("mxfp8", 32, 8),
+            ("nvfp4", 16, 4),
+        ]:
+            for M in [1, 2]:
+                for N in [1, 2, 3, 5, 7]:
+                    with self.subTest(M=M, K=K, N=N, mode=mode):
+                        x = mx.random.normal(shape=(M, K), key=k1) / K**0.5
+                        w = mx.random.normal(shape=(N, K), key=k2) / K**0.5
+                        w_q, scales = mx.quantize(
+                            w, group_size=group_size, bits=bits, mode=mode
+                        )
+                        w_hat = mx.dequantize(
+                            w_q, scales, group_size=group_size, bits=bits, mode=mode
+                        )
+                        y_q = mx.quantized_matmul(
+                            x,
+                            w_q,
+                            scales,
+                            transpose=True,
+                            group_size=group_size,
+                            bits=bits,
+                            mode=mode,
+                        )
+                        y_hat = x @ mx.swapaxes(w_hat, -1, -2)
+                        self.assertEqual(y_q.shape, y_hat.shape)
+                        self.assertLess((y_q - y_hat).abs().max(), 1e-3)
+
     def test_qmv_wide(self):
         # M in [2, vector_limit) routes to qmv_wide -- except K in {64, 128}
         # with power-of-2 bits, which stays on qmv_quad. Check both paths