diff --git a/tests/attention/test_gqa_short_context_sweep_decode_composite_fused.py b/tests/attention/test_gqa_short_context_sweep_decode_composite_fused.py index 1c8c9c0..91b1228 100644 --- a/tests/attention/test_gqa_short_context_sweep_decode_composite_fused.py +++ b/tests/attention/test_gqa_short_context_sweep_decode_composite_fused.py @@ -42,7 +42,7 @@ MODES = [ CONTEXT_LENGTHS = [8 * 1024, 16 * 1024, 32 * 1024, 64 * 1024] CSV_OUT = (Path(__file__).resolve().parents[2] - / "docs" / "sweeps" / "short_context_decode_composite_sweep.csv") + / "docs" / "sweeps" / "short_context_decode_composite_fused_sweep.csv") def _compute_metrics(r, *, mode, kv_per_cube, C, S_kv): diff --git a/tests/attention/test_gqa_short_context_sweep_prefill_composite.py b/tests/attention/test_gqa_short_context_sweep_prefill_composite.py index 699af9f..73108c4 100644 --- a/tests/attention/test_gqa_short_context_sweep_prefill_composite.py +++ b/tests/attention/test_gqa_short_context_sweep_prefill_composite.py @@ -42,7 +42,7 @@ MODES = [ CONTEXT_LENGTHS = [8 * 1024, 16 * 1024, 32 * 1024, 64 * 1024] CSV_OUT = (Path(__file__).resolve().parents[2] - / "docs" / "sweeps" / "short_context_decode_composite_sweep.csv") + / "docs" / "sweeps" / "short_context_prefill_composite_sweep.csv") def _compute_metrics(r, *, mode, kv_per_cube, C, S_kv): diff --git a/tests/attention/test_gqa_short_context_sweep_prefill_composite_fused.py b/tests/attention/test_gqa_short_context_sweep_prefill_composite_fused.py index 9a935ef..d1761df 100644 --- a/tests/attention/test_gqa_short_context_sweep_prefill_composite_fused.py +++ b/tests/attention/test_gqa_short_context_sweep_prefill_composite_fused.py @@ -42,7 +42,7 @@ MODES = [ CONTEXT_LENGTHS = [8 * 1024, 16 * 1024, 32 * 1024, 64 * 1024] CSV_OUT = (Path(__file__).resolve().parents[2] - / "docs" / "sweeps" / "short_context_decode_composite_sweep.csv") + / "docs" / "sweeps" / "short_context_prefill_composite_fused_sweep.csv") def _compute_metrics(r, *, mode, kv_per_cube, C, S_kv):