torch 2.8.0 nightly support

[mhs4670go]

What

Now, torch 2.8.0 nightly version is released. Let's support this version.

There are a few things to resolve.

• Add 2.8 to the if condition.

TICO/tico/utils/convert.py

Line 130 in 536c52b

elif torch.__version__.startswith("2.6") or torch.__version__.startswith("2.7"):

• Ignore UserWarning

There's a user warning about not having GPU. Seems that it can be ignored.

test_add_float (unit_test.utils_test.test_infer.InferSimpleAddTest) ... /home/seongwoo/circle-exir/.venv2/lib/python3.10/site-packages/torch/backends/mkldnn/__init__.py:78: UserWarning: TF32 acceleration on top of oneDNN is available for Intel GPUs. The current Torch version does not have Intel GPU Support. (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:148.)
  torch._C._set_onednn_allow_tf32(_allow_tf32)
/home/seongwoo/circle-exir/.venv2/lib/python3.10/site-packages/torch/backends/mkldnn/__init__.py:78: UserWarning: TF32 acceleration on top of oneDNN is available for Intel GPUs. The current Torch version does not have Intel GPU Support. (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:148.)
  torch._C._set_onednn_allow_tf32(_allow_tf32)

• Support a new operator

It happens with test.modules.net.interpolate.InterpolateDouble.

 NOT SUPPORTED OPERATOR IN GRAPH MODULE
ERROR
test.modules.net.interpolate.InterpolateThreeTimes (pt2_to_circle_test.test_net.test.modules.net.interpolate) ... /home/seongwoo/circle-exir/.venv2/lib/python3.10/site-packages/torch/backends/mkldnn/__init__.py:78: UserWarning: TF32 acceleration on top of oneDNN is available for Intel GPUs. The current Torch version does not have Intel GPU Support. (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:148.)
  torch._C._set_onednn_allow_tf32(_allow_tf32)
/home/seongwoo/circle-exir/.venv2/lib/python3.10/site-packages/torch/backends/mkldnn/__init__.py:78: UserWarning: TF32 acceleration on top of oneDNN is available for Intel GPUs. The current Torch version does not have Intel GPU Support. (Triggered internally at /pytorch/aten/src/ATen/Context.cpp:148.)
  torch._C._set_onednn_allow_tf32(_allow_tf32)
ERROR:circle_exir.utils.convert:NOT SUPPORTED OPERATOR
	(op) upsample_nearest2d.vec
	(trace) File "/home/seongwoo/circle-exir/test/modules/net/interpolate.py", line 34, in forward
    return torch.nn.functional.interpolate(x, scale_factor=3.0, mode="nearest")

[dayo09]

Let me work on this :-D

[dayo09]

Done!