GitLab

* README syntax highlighting

Adds syntax highlighting to README

* Update README.md

remove checkout v2.0.0.post1 from dockerfile

[LayerNorm] Fix typo in ln_api.cpp

unintialized -> uninitialized

Fix compile error with `BOOL_SWITCH`

See also: https://stackoverflow.com/questions/55136414/constexpr-variable-captured-inside-lambda-loses-its-constexpr-ness

Metal FlashAttention

rotary: update cos/sin cache when switching from inference mode

This resolves RuntimeErrors after running evaluation in inference mode:

```
  File "/home/proger/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/home/proger/.local/lib/python3.10/site-packages/flash_attn/modules/mha.py", line 492, in forward
    qkv = self.rotary_emb(qkv)
  File "/home/proger/.local/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl
    return forward_call(*args, **kwargs)
  File "/home/proger/.local/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 229, in forward
    return apply_rotary_emb_qkv_(
  File "/home/proger/.local/lib/python3.10/site-packages/torch/autograd/function.py", line 506, in apply
    return super().apply(*args, **kwargs)  # type: ignore[misc]
RuntimeError: Inference tensors cannot be saved for backward. To work around you can make a clone to get a normal tensor and use it in autograd.
```

I haven't found an easy way to add torch as a build dependency in
pyproject.toml.
If we add torch in pyproject.toml, for some setup it would download a
different version of Pytorch before building.
If we don't add torch, lots of users report they get error when installing.

Fix a bug