[Bug]: Convert operation f16 -> f64 produces incorrect/corrupted values on GPU devices

[zhihaoxu1325]

OpenVINO Version

2025.4.0

Operating System

Ubuntu 20.04 (LTS)

Device used for inference

GPU

Framework

None

Model used

No response

Issue description

When using the GPU plugin, the Convert operation from float16 to float64 produces incorrect values. The CPU plugin produces correct results.

Step-by-step reproduction

Minimal reproducer:

import numpy as np
import openvino as ov
import openvino.runtime.opset13 as ops
from openvino.runtime import Model
from openvino import Type

core = ov.Core()

f16_data = np.array([[4.2421875, 5.85546875]], dtype=np.float16)
const_f16 = ops.constant(f16_data)
convert_to_f64 = ops.convert(const_f16, Type.f64)
model = Model([convert_to_f64], [])

cpu_result = np.array(core.compile_model(model, "CPU")({})[0])
gpu_result = np.array(core.compile_model(model, "GPU")({})[0])

print(f"Input (f16):  {f16_data}")
print(f"CPU (f64):    {cpu_result}")  # Correct: [[4.2421875, 5.85546875]]
print(f"GPU (f64):    {gpu_result}")  # Wrong:   [[0.0, 2.26513672]]

Relevant log output

Input (f16):  [[4.242 5.855]]
CPU (f64):    [[4.2421875  5.85546875]]
GPU (f64):    [[0.         2.26513672]]

Issue submission checklist

• I'm reporting an issue. It's not a question.
• I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• There is reproducer code and related data files such as images, videos, models, etc.

[YuChern-Intel]

Yes, this issue can be reproduced.

I’ll escalate the case to relevant team for further investigation. We will provide an update soon.

[jgespino]

Ref. 179089

[jgespino]

@zhihaoxu1325 GPU plugin does not support f64 data type, please take a look at the supported inference data types.

https://docs.openvino.ai/2025/openvino-workflow/running-inference/inference-devices-and-modes/gpu-device.html#supported-inference-data-types