sycl: fix UE4M3 parsing (#25608)

The NVFP4 quantization format stores a scaling factor for every group of
16 weights, packed into a single UE4M3 byte.

The SYCL GPU code was converting these scale values using the E4M3 path,
but that's *signed*, and these are unsigned values.

ggml/src/ggml-sycl/common.hpp

@@ -1022,9 +1022,20 @@ static T block_reduce(T val, T * shared_vals, int block_size_template) {
 }
 
 static __dpct_inline__ float ggml_sycl_ue4m3_to_fp32(uint8_t x) {
-    const uint32_t bits = x * (x != 0x7F && x != 0xFF);
+    // UE4M3 is unsigned: 4 exp bits (bias 7), 3 mantissa bits, no sign, no NaN.
-    const __nv_fp8_e4m3 xf = *reinterpret_cast<const __nv_fp8_e4m3 *>(&bits);
+    // exp == 0xF is a valid exponent (256-448 range), not NaN.
-    return static_cast<float>(xf) / 2;
+    if (x == 0 || x == 0x7F) {
+        return 0.0f;
+    }
+    const int exp = (x >> 3) & 0xF;
+    const int man = x & 0x7;
+    float raw;
+    if (exp == 0) {
+        raw = man * (1.0f / 8.0f) * sycl::pow(2.0f, -6.0f);
+    } else {
+        raw = (1.0f + man / 8.0f) * sycl::pow(2.0f, (float) exp - 7.0f);
+    }
+    return raw * 0.5f;
 }
 
 #endif // GGML_SYCL_COMMON_HPP