Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
LLM面试常见手撕合集
| Date | Stars |
|---|---|
| 2026-07-31 | 540 |
| 2026-08-06 | 549 |
Today
+9 stars today
This week
— stars this week
This month
— stars this month
Momentum
21.0
growth rate 0.00%/day
# LLM Interview Code **LLM面试常见手撕代码合集** > ps: 本人目前几十场面试仅遇到过 `MHA`, `RoPE`, `RMSNorm`, `BPE`, `InfoNCE`, `DPO`, `工具调用解析`。如有帮助请点个star⭐️~ --- 我去，这么多⭐️，感谢大家～ ## 项目结构 <table> <thead> <tr> <th>目录</th> <th>文件</th> <th>说明</th> </tr> </thead> <tbody> <tr> <td rowspan="4"><strong>Attention</strong></td> <td><a href="./Attention/MHA.ipynb">MHA.ipynb</a></td> <td>多头注意力 (Multi-Head Attention)</td> </tr> <tr> <td><a href="./Attention/GQA.ipynb">GQA.ipynb</a></td> <td>分组查询注意力 (Grouped Query Attention)</td> </tr> <tr> <td><a href="./Attention/MLA.ipynb">MLA.ipynb</a></td> <td>多头潜注意力（Multi-Head Latent Attention）</td> </tr> <tr> <td><a href="./Attention/MHA_kvcache.ipynb">MHA_kvcache.ipynb</a></td> <td>带KV cache的注意力</td> </tr> <tr> <td><a href="./Attention/mask.ipynb">mask.ipynb</a></td> <td>注意力掩码</td> </tr> <tr> <td rowspan="5"><strong>Components</strong></td> <td><a href="./Components/Linear.ipynb">Linear.ipynb</a></td> <td>线性层</td> </tr> <tr> <td><a href="./Components/BPE.ipynb">BPE.ipynb</a></td> <td>Byte Pair Encoding</td> </tr> <tr> <td><a href="./Components/LoRA.ipynb">LoRA.ipynb</a></td> <td>LoRA Linear 层</td> </tr> <tr> <td><a href="./Components/RoPE.ipynb">RoPE.ipynb</a></td> <td>旋转位置编码</td> </tr> <tr> <td><a href="./Components/SwiGLU.ipynb">SwiGLU.ipynb</a></td> <td>SwiGLU 激活函数</td> </tr> <tr> <td rowspan="2"><strong>Norm</strong></td> <td><a href="./Norm/LayerNorm.ipynb">LayerNorm.ipynb</a></td> <td>层归一化</td> </tr> <tr> <td><a href="./Norm/RMSNorm.ipynb">RMSNorm.ipynb</a></td> <td>RMS归一化</td> </tr> <tr> <td rowspan="7"><strong>Functional</strong></td> <td><a href="./Functional/activation_fun.ipynb">activation_fun.ipynb</a></td> <td>激活函数</td> </tr> <tr> <td><a href="./Functional/CE.ipynb">CE.ipynb</a></td> <td>交叉熵损失</td> </tr> <tr> <td><a href="./Functional/InfoNCE.ipynb">InfoNCE.ipynb</a></td> <td>InfoNCE损失</td> </tr> <tr> <td><a href="./Functional/quantize.ipynb">quantize.ipynb</a></td> <td>量化</td> </tr> <tr> <td><a href="./Functional/sft.ipynb">sft.ipynb</a></td> <td>SFT损失</td> </tr> <tr> <td><a href="./Functional/sample.ipynb">sample.ipynb</a></td> <td>采样方法</td> </tr> <tr> <td><a href="./Functional/tool_function.ipynb">tool_function.ipynb</a></td> <td>Tool Calling 流式解析</td> </tr> <tr> <td rowspan="5"><strong>RL</strong></td> <td><a href="./RL/DPO.ipynb">DPO.ipynb</a></td> <td>DPO损失</td> </tr> <tr> <td><a href="./RL/GRPO.ipynb">GRPO.ipynb</a></td> <td>GRPO损失</td> </tr> <tr> <td><a href="./RL/DAPO.ipynb">DAPO.ipynb</a></td> <td>DAPO损失</td> </tr> <tr> <td><a href="./RL/GSPO.ipynb">GSPO.ipynb</a></td> <td>GSPO损失</td> </tr> <tr> <td><a href="./RL/KL.ipynb">KL.ipynb</a></td> <td>KL散度</td> </tr> </tbody> </table>
Excerpt of 2,655 characters
Read on GitHub
12
Would you bet a product on this? Bounded 0–100 and slow moving.
matched fp:d9065dc4b9f58785, topic:llm