[fix] baseline4 copy_blocks 崩溃和 ex_engine import 路径两个接线没接对
This commit is contained in:
@@ -2282,5 +2282,4 @@ class PagedAttention:
|
|||||||
) -> None:
|
) -> None:
|
||||||
key_caches = [kv_cache[0] for kv_cache in kv_caches]
|
key_caches = [kv_cache[0] for kv_cache in kv_caches]
|
||||||
value_caches = [kv_cache[1] for kv_cache in kv_caches]
|
value_caches = [kv_cache[1] for kv_cache in kv_caches]
|
||||||
import ixformer.functions as ixf_F
|
ops.copy_blocks(key_caches, value_caches, src_to_dists)
|
||||||
ixf_F.vllm_copy_cache(key_caches, value_caches, src_to_dists)
|
|
||||||
@@ -247,10 +247,19 @@ INIT_EOF
|
|||||||
import logging
|
import logging
|
||||||
_logger = logging.getLogger("ix_startup_patch")
|
_logger = logging.getLogger("ix_startup_patch")
|
||||||
def apply():
|
def apply():
|
||||||
|
import sys, os
|
||||||
|
# Ensure ex_engine is importable from both locations
|
||||||
|
for p in ["/workspace/qwen3_6_scripts/ex_engine/..",
|
||||||
|
"/workspace/qwen3_6_scripts",
|
||||||
|
"/workspace/ex_engine/..",
|
||||||
|
"/workspace"]:
|
||||||
|
rp = os.path.realpath(p)
|
||||||
|
if os.path.isdir(rp) and rp not in sys.path:
|
||||||
|
sys.path.insert(0, rp)
|
||||||
n = 0
|
n = 0
|
||||||
# 1. ix_full_bridge patches (rms_norm, silu_and_mul, linear)
|
# 1. ix_full_bridge patches (rms_norm, silu_and_mul, linear)
|
||||||
try:
|
try:
|
||||||
from vllm.ex_engine.python.patch_vllm_ops import apply_all_patches
|
from ex_engine.python.patch_vllm_ops import apply_all_patches
|
||||||
k = apply_all_patches()
|
k = apply_all_patches()
|
||||||
n += k
|
n += k
|
||||||
if k > 0:
|
if k > 0:
|
||||||
@@ -259,7 +268,7 @@ def apply():
|
|||||||
_logger.warning("ix_startup_patch: bridge patches failed: %s", e)
|
_logger.warning("ix_startup_patch: bridge patches failed: %s", e)
|
||||||
# 2. xllm kernel patches (topk_softmax, norm, activation, rope, cache)
|
# 2. xllm kernel patches (topk_softmax, norm, activation, rope, cache)
|
||||||
try:
|
try:
|
||||||
from vllm.ex_engine.python.patch_vllm_hot_path import apply as apply_hot
|
from ex_engine.python.patch_vllm_hot_path import apply as apply_hot
|
||||||
k = apply_hot(strict=False)
|
k = apply_hot(strict=False)
|
||||||
n += k
|
n += k
|
||||||
if k > 0:
|
if k > 0:
|
||||||
|
|||||||
Reference in New Issue
Block a user