Revert "Enable memory saver for hybrid model" (#12648)

This commit is contained in:
Baizhou Zhang
2025-11-04 16:22:06 -08:00
committed by GitHub
parent 887742a1e7
commit d22d044734
3 changed files with 8 additions and 15 deletions
-2
View File
@@ -42,7 +42,6 @@ class TestMamba(unittest.TestCase):
full_attention_layer_ids=full_attention_layer_ids,
enable_kvcache_transpose=False,
device=device,
enable_memory_saver=False,
mamba_pool=None,
)
assert pool._transfer_full_attention_id(global_interval - 1) == 0
@@ -175,7 +174,6 @@ class TestMamba(unittest.TestCase):
full_attention_layer_ids=full_attention_layer_ids,
enable_kvcache_transpose=False,
device=device,
enable_memory_saver=False,
mamba_pool=req_to_token_pool.mamba_pool,
)