Skip to content

Actions: ztxz16/fastllm

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
465 workflow runs
465 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
优化 Hopper BF16 解码融合算子 Action Build #791: Commit a2ff521 pushed by ztxz16
2m 53s master
master
2m 53s
优化 Hopper BF16 线性投影 Action Build #790: Commit fab2b14 pushed by ztxz16
3m 55s master
master
3m 55s
优化 Hopper FP16 输出投影 Action Build #789: Commit a2e20c8 pushed by ztxz16
4m 11s master
master
4m 11s
修正 decode 动态分流估算并流水上传专家 Action Build #788: Commit 291974d pushed by ztxz16
4m 10s master
master
4m 10s
补充 Naive TP 历史复用回归与说明 Action Build #787: Commit e2edc8a pushed by ztxz16
4m 0s master
master
4m 0s
共享 Attention 输出合并内核 Action Build #786: Commit 7f739ad pushed by ztxz16
4m 7s master
master
4m 7s
修复 GLM TP 推测解码的纯 CPU 编译 Action Build #785: Commit cc849fe pushed by ztxz16
4m 57s master
master
4m 57s
支持草稿独立 TP 与上下文批量投影 Action Build #784: Commit 02f3af7 pushed by ztxz16
2m 56s master
master
2m 56s
复用 QSA 预热锁页内存降低首请求延迟 Action Build #783: Commit 4a1d606 pushed by ztxz16
4m 12s master
master
4m 12s
扩展原生权重的 TP 流式加载 Action Build #782: Commit 794fe05 pushed by ztxz16
5m 52s master
master
5m 52s
优化CUDA预填充计算与临时显存 Action Build #781: Commit 16209a4 pushed by ztxz16
4m 26s master
master
4m 26s
预读并缓存 Qwen4 PLE 行 Action Build #780: Commit 8eb2514 pushed by ztxz16
7m 26s master
master
7m 26s
优化 IQ4_XS 原始布局 CPU 计算 Action Build #779: Commit b576532 pushed by ztxz16
7m 25s master
master
7m 25s
降低 Qwen4 TP 主机内存占用 Action Build #778: Commit 74037df pushed by ztxz16
6m 52s master
master
6m 52s
补充未发布版本更新说明 Action Build #777: Commit f026aa5 pushed by ztxz16
5m 31s master
master
5m 31s
支持临时专家直接读取 NUMA 量化布局 Action Build #776: Commit bc4b6b9 pushed by ztxz16
7m 34s master
master
7m 34s
优化 DFlash 双请求验证与状态恢复 Action Build #775: Commit 97d7f21 pushed by ztxz16
5m 33s master
master
5m 33s
优化MLA批量解码与调度开销 Action Build #774: Commit 129dcff pushed by ztxz16
4m 31s master
master
4m 31s
实现 prefill 缓存填充与专家流水线 Action Build #773: Commit 677d689 pushed by ztxz16
7m 2s master
master
7m 2s
复用 NUMA GGUF 权重降低内存占用 Action Build #772: Commit 6812932 pushed by ztxz16
7m 5s master
master
7m 5s
补齐多卡 MoE 动态上传分流 Action Build #771: Commit 5a0ee98 pushed by ztxz16
5m 29s master
master
5m 29s
修正多卡专家并行测试断言 Action Build #770: Commit 2dbbb3c pushed by ztxz16
7m 9s master
master
7m 9s
接入 GLM GGUF 频率缓存 Action Build #769: Commit 924643f pushed by ztxz16
7m 14s master
master
7m 14s
修复 SM75 的 DSA 内核链接 Action Build #768: Commit 4e36d8e pushed by ztxz16
4m 32s master
master
4m 32s
减少 DFlash 验证状态备份 Action Build #767: Commit 9037a49 pushed by ztxz16
7m 10s master
master
7m 10s