诊断日志新增 handle-read-packet / handle-engine / handle-selection / handle-worker-push / handle-async-query / handle-candidates 六类分段 (>=8ms 才输出),用于定位 server 按键处理偶发数百毫秒卡顿——昨晚 request=628 卡 344ms 导致空格排队超 TSF 300ms 提交预算、整会话重置 吞掉 "juede"。纯观测改动,零行为变化;.gitignore 忽略本地测试暂存 目录 .tmp-test/。
38 lines
1.2 KiB
Plaintext
38 lines
1.2 KiB
Plaintext
.vs/
|
|
out/
|
|
build/
|
|
Release/
|
|
build64/
|
|
build64-release/
|
|
build32/
|
|
build32-release/
|
|
draft/
|
|
*.log
|
|
CMakeSettings.json
|
|
log.txt
|
|
.workbuddy/
|
|
|
|
# Release staging. The dictionary release is the only product input that still comes from outside
|
|
# this repository, and both the release workflow and a local release rehearsal drop it beside the
|
|
# components under this name.
|
|
/MetasequoiaImeDict/
|
|
|
|
# 语言模型:来源锁和授权声明进仓库,下载的语料、转换工具的构建目录和编出来的 sc.lm
|
|
# 不进 —— 它们都能由 scripts/build-language-model.ps1 从 lock.json 重新得到。
|
|
/language-model/cache/
|
|
/language-model/sc.lm
|
|
/language-model/sc.lm.incoming
|
|
|
|
# 神经整句模型:lock.json 和 NOTICE.md 进仓库,下载下来的 safetensors 不进 ——
|
|
# 由 scripts/fetch-neural-model.ps1 按 lock.json 从 release 重新取到并校验 sha256。
|
|
/neural-model/*.safetensors
|
|
__pycache__/
|
|
|
|
# 外部模型包:models/ 下是按需下载的大文件(如万象 .gram 400MB),来源与摘要
|
|
# 记在调研文档与锁文件里,不进仓库。
|
|
/models/
|
|
|
|
# 本地测试数据暂存(照 scripts/ci/test-server.ps1 的布局手工准备的)
|
|
/stage/
|
|
/.tmp-test/
|