dylanyunlong
0 Followers · 0 Following
  • Joined on 2026-08-06
Repositories
6
 Projects Public Activity Starred Repositories
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 22:01:54 +08:00
11a8f3832a fix(vision): 搬运xllm compute_qwen2_vision_attention_cuda替换推理版本
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 21:19:16 +08:00
c6b9ee93e9 fix: total_mem → total_memory
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 21:17:45 +08:00
2f19498ae6 fix: 去掉einops依赖 + 修dist_utils import路径 + 真机验证脚本
a7bedb33ee fix(CRITICAL): patch qwen2_vl vision attention — bypass xops varlen_fwd on BI-V100
5b2b8dcc2a fix(CRITICAL): route ALL prefill through sdpa_fallback — ixformer varlen_fwd incompatible with FwOp 20-arg signature
Compare 3 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 19:54:09 +08:00
aadff65af6 fix(build): 恢复26e6cb40完全一样的Dockerfile结构——5个独立RUN层
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 19:45:38 +08:00
da7d3af56e fix(CRITICAL): launch_server.py覆盖所有vllm文件+运行xformers patch
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 19:14:05 +08:00
79bb35de9f fix(build): patch_ops.sh加cd "$(dirname "$0")" + VLLM_ROOT fallback搜索
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 19:10:11 +08:00
ca6dcc81d2 fix(build): 单RUN + 每步||true——确保Docker build不会因任何编译失败而中断
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 18:52:40 +08:00
d9064550b2 fix(CRITICAL): launch_server.py用execvp启动api_server——import *不触发__main__
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 18:28:14 +08:00
ba0f67e79e fix(runtime): launch_server.py强制覆盖所有vllm路径后启动
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 18:16:00 +08:00
075b5fa18e fix(runtime): 去掉--reasoning-parser qwen3——基础镜像不识别
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 18:02:00 +08:00
4702505bf9 fix(build): Dockerfile还原到26e6cb4结构——3 COPY + 5 RUN
c152bd5a89 feat: ex_factor_0.so ctypes桥接 + ex_engine package部署
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:50:19 +08:00
3a5cc2a589 fix(CRITICAL): moe_topk_softmax用importlib加载+patch_ops编译部署
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:48:56 +08:00
f4d4219280 fix(bridge): ix_moe_bridge链接加--unresolved-symbols + probe脚本
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:47:37 +08:00
1cf4a34d17 fix: link torch_python for pybind11 symbols
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:44:39 +08:00
7e8605248a feat(GDN): 系统设计 — flash_qla_sm70 CUDA kernel + threshold=20.0 gate clamp
6d063fb610 feat: build_moe_topk.sh — compile moe_topk_softmax_v3.cu with correct TORCH_EXTENSION_NAME
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:38:21 +08:00
490ff98ad6 fix(build): 消除COPY ./vllm_overrides — vendor_overrides预置到qwen3_6_scripts/
97d9842180 fix(CRITICAL): bridge build delayed binding + Docker tolerance + improved preload
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:28:52 +08:00
1d5856f4a9 fix: Dockerfile Step 7 VLLM_ROOT污染 — vllm INFO日志混入stdout
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 17:23:42 +08:00
589b91d653 test: 真机build验证脚本 — 模拟Dockerfile 8步
2f5be7d635 fix: GDN NaN clamp (7 sites) + Dockerfile Step 4 tolerance
Compare 2 commits »
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 16:22:53 +08:00
ed8bdf8714 fix(CRITICAL): merge 26e6cb40 build pipeline + HEAD features — fix docker build
dylanyunlong pushed to main at dylanyunlong/project_6 2026-08-11 15:51:33 +08:00
18b52c3db0 debug: list ixformer.functions available APIs