← All skills

pytorch_dynamic_loss_weighting

100 / 100 · Low risk

Fine to try.

By ECNU-ICALK/AutoSkill (GitHub) · other · 613 stars · scanned at commit 94c47ca

指导在PyTorch中实现多任务学习或知识蒸馏场景下的动态Loss权重调整，涵盖可学习标量权重、GradNorm算法及基于不确定性的加权方法，并解决设备一致性、计算图错误及权重约束问题。

Works with any agent that supports SKILL.md skills.

What this skill can do

We found nothing in the files we read that matches our rules. That does not prove the skill is safe: see the limits of static scanning.

Install

Paste this into your terminal. It downloads exactly the version we scanned, not whatever is newest.

if [ -e ~/.agents/skills/'pytorch-dynamic-loss-weighting' ] || [ -L ~/.agents/skills/'pytorch-dynamic-loss-weighting' ]; then echo 'A skill folder named pytorch-dynamic-loss-weighting already exists — not replacing it' >&2; false; else d=$(mktemp -d) && git clone --quiet --filter=blob:none --no-checkout 'https://github.com/ECNU-ICALK/AutoSkill' "$d" && git -C "$d" checkout --quiet '94c47ca488d4ba4117d20272e66d49b9877e68cf' -- 'SkillBank/ConvSkill/chinese_gpt4_8/pytorch_dynamic_loss_weighting' && mkdir -p ~/.agents/skills && cp -R "$d"/'SkillBank/ConvSkill/chinese_gpt4_8/pytorch_dynamic_loss_weighting' ~/.agents/skills/'pytorch-dynamic-loss-weighting' && rm -rf "$d"; fi

Codex reads ~/.agents/skills directly. For Claude Code, Cursor, Copilot, Gemini CLI, Antigravity or OpenCode, also link it. For Claude Code: ln -s ~/.agents/skills/'pytorch-dynamic-loss-weighting' ~/.claude/skills/'pytorch-dynamic-loss-weighting' (other agents: ~/.cursor/skills, ~/.copilot/skills, ~/.gemini/skills, ~/.gemini/config/skills, ~/.config/opencode/skills). Or install through the finder skill, which does both.

Badge for your README

The badge for ECNU-ICALK/AutoSkill shows the score of its lowest-scoring skill.

[![SkillTrust](https://zalavijaysinh.github.io/skilltrust/badges/ECNU-ICALK/AutoSkill.svg)](https://zalavijaysinh.github.io/skilltrust/repos/ECNU-ICALK/AutoSkill/)

Score history