PTQ-Level Efficiency, QAT-Level Accuracy | 训练后量化的效率,量化感知训练的精度
modified repack fp16-bf16 route tests
jijiazbe91de74c6a5
diff --git a/auto_round_extension/ark/test/test_ark_cpu_mixed_bestla_sdpa.py b/auto_round_extension/ark/test/test_ark_cpu_mixed_bestla_sdpa.pyindex 79cd352a9..f15c925c2 100644--- a/auto_round_extension/ark/test/test_ark_cpu_mixed_bestla_sdpa.py+++ b/auto_round_extension/ark/test/test_ark_cpu_mixed_bestla_sdpa.py@@ -53,6 +53,7 @@ def _kernel_view_kv(k, v, seq_q):if (k.dtype == torch.float16and seq_q > 1+ and auto_round_kernel._cpu_public_packed_kv_available()and getattr(auto_round_kernel.cpu_lib, "ARK_CPU_SDPA_BUILD_HAS_BF16_ROUTE", False)):k = k.to(dtype=torch.bfloat16)@@ -74,7 +75,7 @@ def test_mixed_dtype_sdpa_routes_to_mixed_path():out = auto_round_kernel.sdpa(q, k, v, scale=scale)- atol, rtol = _TOL[torch.float16]+ atol, rtol = _TOL[k_ref.dtype]assert out.dtype == torch.float32torch.testing.assert_close(out, expected, atol=atol, rtol=rtol)@@ -101,7 +102,7 @@ def test_bestla_mixed_sdpa_matches_torch(kv_dtype, is_causal, layout):except (RuntimeError, ValueError) as exc:pytest.skip(f"BestLA mixed path unavailable on this ISA/runtime: {exc}")- atol, rtol = _TOL[kv_dtype]+ atol, rtol = _TOL[k_ref.dtype]assert actual.dtype == torch.float32torch.testing.assert_close(_to_hnd(actual, layout), expected_hnd, atol=atol, rtol=rtol)@@ -127,7 +128,7 @@ def test_bestla_mixed_sdpa_gqa_ratio(kv_dtype, gqa_ratio):except (RuntimeError, ValueError) as exc:pytest.skip(f"BestLA mixed path unavailable on this ISA/runtime: {exc}")- atol, rtol = _TOL[kv_dtype]+ atol, rtol = _TOL[k_ref.dtype]assert actual.dtype == torch.float32torch.testing.assert_close(actual, expected, atol=atol, rtol=rtol)@@ -201,7 +202,7 @@ def test_mixed_batched_gqa_prefill_runs_repeatedly(kv_dtype):for _ in range(20):actual = _mixed_sdpa(q, k, v, scale, True, "HND")- atol, rtol = _TOL[kv_dtype]+ atol, rtol = _TOL[k_ref.dtype]assert actual.dtype == torch.float32torch.testing.assert_close(actual, expected, atol=atol, rtol=rtol)@@ -250,6 +251,6 @@ def test_mixed_llm_shape_accuracy(label, batch, heads_q, heads_kv, head_dim, seqroute == auto_round_kernel.cpu_lib.ARK_CPU_SDPA_ROUTE_MIXED_RAW), f"{label}: expected mixed route, got {route}"- atol, rtol = _TOL[kv_dtype]+ atol, rtol = _TOL[k_ref.dtype]assert actual.dtype == torch.float32torch.testing.assert_close(actual, expected, atol=atol, rtol=rtol)