Update PyTorch installation instructions; Fix import and dimensional bugs

[Met4physics]

1. Torch 2.8.0 has been released as a stable version, and the current setup commands will raise errors.
2. FA3 is incorrectly imported after commit Initial AMD MI300X Support via. AITER #10 . This fix also answers issue flash_attn_3 not imported #19 .
3. There is a dimensional bug will cause error:

Traceback (most recent call last):
  File "/storage/lmh/reward/flux-fast/run_benchmark.py", line 85, in <module>
    main(args)
    ~~~~^^^^^^
  File "/storage/lmh/reward/flux-fast/run_benchmark.py", line 34, in main
    image = pipeline(
            ~~~~~~~~^
        prompt=args.prompt,
        ^^^^^^^^^^^^^^^^^^^
    ...<2 lines>...
        **_determine_pipe_call_kwargs(args)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ).images[0]
    ^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/pipelines/flux/pipeline_flux.py", line 919, in __call__
    noise_pred = self.transformer(
                 ~~~~~~~~~~~~~~~~^
        hidden_states=latents,
        ^^^^^^^^^^^^^^^^^^^^^^
    ...<7 lines>...
        return_dict=False,
        ^^^^^^^^^^^^^^^^^^
    )[0]
    ^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
    return forward_call(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/models/transformers/transformer_flux.py", line 733, in forward
    encoder_hidden_states, hidden_states = block(
                                           ~~~~~^
        hidden_states=hidden_states,
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    ...<3 lines>...
        joint_attention_kwargs=joint_attention_kwargs,
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    )
    ^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
    return forward_call(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/models/transformers/transformer_flux.py", line 456, in forward
    attention_outputs = self.attn(
        hidden_states=norm_hidden_states,
    ...<2 lines>...
        **joint_attention_kwargs,
    )
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
    return forward_call(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/models/transformers/transformer_flux.py", line 343, in forward
    return self.processor(self, hidden_states, encoder_hidden_states, attention_mask, image_rotary_emb, **kwargs)
           ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/storage/lmh/reward/flux-fast/utils/pipeline_utils.py", line 175, in __call__
    value.transpose(1, 2))[0].transpose(1, 2)
                              ~~~~~~~~~^^^^^^
IndexError: Dimension out of range (expected to be in range of [-2, 1], but got 2)

4. Current hard-coded operation will cause dimentional error when using compile+fa3.

   flux-fast/utils/pipeline_utils.py

   Line 63 in 900ea5b

   return outputs[0]

Traceback (most recent call last):
  File "/storage/lmh/reward/flux-fast/run_benchmark.py", line 85, in <module>
    main(args)
    ~~~~^^^^^^
  File "/storage/lmh/reward/flux-fast/run_benchmark.py", line 27, in main
    pipeline = load_pipeline(args)
  File "/storage/lmh/reward/flux-fast/utils/pipeline_utils.py", line 507, in load_pipeline
    pipeline = optimize(pipeline, args)
  File "/storage/lmh/reward/flux-fast/utils/pipeline_utils.py", line 479, in optimize
    pipeline = use_compile(pipeline)
  File "/storage/lmh/reward/flux-fast/utils/pipeline_utils.py", line 292, in use_compile
    pipeline(**input_kwargs).images[0]
    ~~~~~~~~^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/utils/_contextlib.py", line 120, in decorate_context
    return func(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/pipelines/flux/pipeline_flux.py", line 919, in __call__
    noise_pred = self.transformer(
                 ~~~~~~~~~~~~~~~~^
        hidden_states=latents,
        ^^^^^^^^^^^^^^^^^^^^^^
    ...<7 lines>...
        return_dict=False,
        ^^^^^^^^^^^^^^^^^^
    )[0]
    ^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_dynamo/eval_frame.py", line 375, in __call__
    return super().__call__(*args, **kwargs)
           ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
    return forward_call(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_dynamo/eval_frame.py", line 736, in compile_wrapper
    return fn(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1773, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
           ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/nn/modules/module.py", line 1784, in _call_impl
    return forward_call(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/diffusers/models/transformers/transformer_flux.py", line 631, in forward
    def forward(
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_dynamo/eval_frame.py", line 929, in _fn
    return fn(*args, **kwargs)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_functorch/aot_autograd.py", line 1241, in forward
    return compiled_fn(full_args)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 384, in runtime_wrapper
    all_outs = call_func_at_runtime_with_args(
        compiled_fn, args, disable_amp=disable_amp, steal_args=True
    )
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_functorch/_aot_autograd/utils.py", line 126, in call_func_at_runtime_with_args
    out = normalize_as_list(f(args))
                            ~^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 750, in inner_fn
    outs = compiled_fn(args)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_functorch/_aot_autograd/runtime_wrappers.py", line 556, in wrapper
    return compiled_fn(runtime_args)
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_inductor/output_code.py", line 584, in __call__
    return self.current_callable(inputs)
           ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^
  File "/storage/lmh/miniconda3/envs/fastflux/lib/python3.13/site-packages/torch/_inductor/utils.py", line 2716, in run
    out = model(new_inputs)
  File "/tmp/torchinductor_lmh/de/cde2egsropmjrue2ojtrrndkfnmoyj5jx2fi5g3lefdapzq2xeo4.py", line 4638, in call
    assert_size_stride(buf43, (1, 4608, 24, 128), (14155776, 3072, 128, 1), 'torch.ops.flash.flash_attn_func.default')
    ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: wrong number of dimensions3 for op: torch.ops.flash.flash_attn_func.default