master

lulina1xlite: refact mla_v2 and remove mla_v3

a2bf54ea创建于 14 天前历史提交

文件	最后提交记录	最后更新时间
e2e	xlite: ci: Add benchmark report header and command used Signed-off-by: Sijie Fu <fusijie@huawei.com>	15 天前
funcs	xlite: renormalize line endings Git index had some CRLF files which are checkout on Unix as-is. This commit runs git add --renormalize . command and unifies line endings as LF. Current line endings in repo and in working space can be seen via git ls-files --eol command. Having all files in LF format in index allows both Unix and Windows work correctly with default git settings. Signed-off-by: Igor Prusov <prusov.igor@huawei.com>	2 个月前
kernels	xlite: refact mla_v2 and remove mla_v3 Signed-off-by: lulina1 <lina.lulina@huawei.com>	14 天前
models	xlite: adapt torch backend for glm-5.1 w4a8 test model	18 天前
performance	xlite: Add tools to process swizzle measurements Signed-off-by: Igor Prusov <prusov.igor@huawei.com>	3 个月前
cpp_check.py	xlite: tests: parallelize C++ static checks Run per-file copyright, header-guard, clang-format, and clang-tidy checks concurrently via ThreadPoolExecutor. Check helpers now return error lists instead of mutating shared state, and add a --jobs flag. Signed-off-by: Igor Prusov <prusov.igor@huawei.com> Co-authored-by: Deepseek V4 Pro	1 个月前
generate.py	xlite: fix topk result sharing when skipped with nullptr; replace indexer_skip_layers with index_full_mask for DSA top-k sharing; add the original config.json path to model args Signed-off-by: Sijie Fu <fusijie@huawei.com>	26 天前
pr_ci_bot.py	xlite: fix PR CI bot only fetching earliest comments Co-Authored-By: GLM-5.2 Signed-off-by: lulina1 <lina.lulina@huawei.com>	21 天前
python_check.py	xlite: fix quant grammar problems	5 个月前
run.sh	xlite: adapt torch backend for glm-5.1 w4a8 test model	18 天前
run_accuracy.sh	xlite: only use torchrun when tp size > 1 in testcase Signed-off-by: lulina1 <lina.lulina@huawei.com>	3 个月前
run_perf.sh	xlite: remove private experts int8 quantization Signed-off-by: lulina1 <lina.lulina@huawei.com>	3 个月前
run_static_checks.py	xlite: tests: parallelize C++ static checks Run per-file copyright, header-guard, clang-format, and clang-tidy checks concurrently via ThreadPoolExecutor. Check helpers now return error lists instead of mutating shared state, and add a --jobs flag. Signed-off-by: Igor Prusov <prusov.igor@huawei.com> Co-authored-by: Deepseek V4 Pro	1 个月前