Update python API of activation, topk, norm and rope and remove vllm dependency (#6614)

Co-authored-by: Wu, Chunyuan <chunyuan.wu@intel.com>
Co-authored-by: jianan-gu <jianan.gu@intel.com>
Co-authored-by: sdp <sdp@gnr799219.jf.intel.com>
This commit is contained in:
YanbingJiang
2025-06-18 13:11:50 +08:00
committed by GitHub
parent e56685ac1b
commit 094c116f7d
23 changed files with 270 additions and 56 deletions

View File

@@ -71,7 +71,7 @@ class TestSharedExpert(CustomTestCase):
)
atol = rtol = precision[ref.dtype]
self.assertTrue(torch.allclose(ref, res, atol=atol, rtol=rtol))
torch.testing.assert_close(ref, res, atol=atol, rtol=rtol)
def test_bf16_shared_expert(self):
for params in itertools.product(
@@ -129,7 +129,7 @@ class TestSharedExpert(CustomTestCase):
)
atol = rtol = precision[ref2.dtype]
self.assertTrue(torch.allclose(ref2, res2, atol=atol, rtol=rtol))
torch.testing.assert_close(ref2, res2, atol=atol, rtol=rtol)
def test_int8_shared_expert(self):
for params in itertools.product(
@@ -199,7 +199,7 @@ class TestSharedExpert(CustomTestCase):
)
atol = rtol = precision[ref_out.dtype]
self.assertTrue(torch.allclose(ref_out, out, atol=atol, rtol=rtol))
torch.testing.assert_close(ref_out, out, atol=atol, rtol=rtol)
def test_fp8_shared_expert(self):
for params in itertools.product(