Enhanced Transformer with Rotary Position Embedding. Derived from: https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/ transformers/rope/__init__.py. MIT License: https://github.com/labmlai/annotated_deep_learning_paper_implementati
(
self, seq_len: int, n_elem: int, dtype: torch.dtype, device: torch.device, base: int = 10000
)
| 101 | self.rope_ratio = rope_ratio |
| 102 | |
| 103 | def forward_impl( |
| 104 | self, seq_len: int, n_elem: int, dtype: torch.dtype, device: torch.device, base: int = 10000 |
| 105 | ): |
| 106 | """Enhanced Transformer with Rotary Position Embedding. |
| 107 | |
| 108 | Derived from: https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/labml_nn/ |
| 109 | transformers/rope/__init__.py. MIT License: |
| 110 | https://github.com/labmlai/annotated_deep_learning_paper_implementations/blob/master/license. |
| 111 | """ |
| 112 | # $\Theta = {\theta_i = 10000^{\frac{2(i-1)}{d}}, i \in [1, 2, ..., \frac{d}{2}]}$ |
| 113 | |
| 114 | base = base * self.rope_ratio |
| 115 | theta = 1.0 / (base ** (torch.arange(0, n_elem, 2, dtype=torch.float, device=device) / n_elem)) |
| 116 | |
| 117 | # Create position indexes `[0, 1, ..., seq_len - 1]` |
| 118 | seq_idx = torch.arange(seq_len, dtype=torch.float, device=device) |
| 119 | |
| 120 | # Calculate the product of position index and $\theta_i$ |
| 121 | idx_theta = torch.outer(seq_idx, theta).float() |
| 122 | |
| 123 | cache = torch.stack([torch.cos(idx_theta), torch.sin(idx_theta)], dim=-1) |
| 124 | |
| 125 | # this is to mimic the behaviour of complex32, else we will get different results |
| 126 | if dtype in (torch.float16, torch.bfloat16, torch.int8): |
| 127 | cache = cache.bfloat16() if dtype == torch.bfloat16 else cache.half() |
| 128 | return cache |
| 129 | |
| 130 | def forward(self, max_seq_len, offset=0): |
| 131 | return self.forward_impl( |