Remove V0 attention backends (#25351)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2025-09-21 16:03:28 -07:00
committed by GitHub
parent af7dfb0d1a
commit bc6e542d9f
28 changed files with 143 additions and 7376 deletions
@@ -44,7 +44,7 @@ class model_aware_kv_ops_helper:
# When VLLM_MLA_DISABLE=1, standard FA is used instead, leading
# to a kv_cache shape of [2, num_blks, blk_size,
# num_key_value_heads / tp, qk_nope_head_dim + qk_rope_head_dim].
# For more details, see vllm/attention/backends/mla/common.py.
# For more details, see vllm/v1/attention/backends/mla/common.py.
if self.is_deepseek_mla and self.use_mla_opt:
head_size = model_config.kv_lora_rank + \
model_config.qk_rope_head_dim