File tree Expand file tree Collapse file tree
Expand file tree Collapse file tree Original file line number Diff line number Diff line change 2323
2424class KVCache (nn .Module ):
2525 """
26- MLX-optimized KV cache with ExecutorTorch llama KVCache interface.
26+ MLX-optimized KV cache with ExecuTorch llama KVCache interface.
2727
2828 This class follows the same interface as examples/models/llama/attention.py KVCache,
2929 making it a drop-in replacement, but uses the mlx::kv_cache_update op internally
Original file line number Diff line number Diff line change 77# LICENSE file in the root directory of this source tree.
88
99"""
10- MLX-optimized attention for ExecutorTorch 's Llama attention registry.
10+ MLX-optimized attention for ExecuTorch 's Llama attention registry.
1111
1212Registers an "mlx" attention type that uses mlx::kv_cache_update and
1313mlx::custom_sdpa for efficient execution on Apple Silicon.
Original file line number Diff line number Diff line change @@ -1810,7 +1810,7 @@ class KVCacheModel(nn.Module):
18101810 """
18111811 Test model wrapping KVCache from cache.py.
18121812
1813- This tests the ExecutorTorch llama KVCache-compatible interface that uses
1813+ This tests the ExecuTorch llama KVCache-compatible interface that uses
18141814 the mlx::kv_cache_update op internally.
18151815 """
18161816
@@ -1845,7 +1845,7 @@ def forward(
18451845@register_test
18461846class KVCacheTest (OpTestCase ):
18471847 """
1848- Test case for MLX KVCache with ExecutorTorch llama KVCache interface.
1848+ Test case for MLX KVCache with ExecuTorch llama KVCache interface.
18491849
18501850 This verifies that KVCache:
18511851 1. Accepts the ET llama KVCache update interface
You can’t perform that action at this time.
0 commit comments