摘要: ![](https://img2023.cnblogs.com/blog/1898939/202309/1898939-20230904182224084-1937412113.png) ```py class ChatGLMForConditionalGeneration(ChatGLMPreTr 阅读全文
posted @ 2023-09-04 18:22 绝不原创的飞龙 阅读(151) 评论(0) 推荐(0)
摘要: ![](https://img2023.cnblogs.com/blog/1898939/202309/1898939-20230904181857291-366474134.png) ```py # 完整的 GLM 模型,包括嵌入层、编码器、输出层 class ChatGLMModel(ChatG 阅读全文
posted @ 2023-09-04 18:19 绝不原创的飞龙 阅读(233) 评论(0) 推荐(0)
摘要: ![](https://img2023.cnblogs.com/blog/1898939/202309/1898939-20230904181352733-216543774.png) ```py # 编码器模块,包含所有 GLM 块 class GLMTransformer(torch.nn.Mo 阅读全文
posted @ 2023-09-04 18:14 绝不原创的飞龙 阅读(147) 评论(0) 推荐(0)
摘要: ![](https://img2023.cnblogs.com/blog/1898939/202309/1898939-20230904180938085-614439600.png) ```py # GLM 块包括注意力层、FFN层和之间的残差 class GLMBlock(torch.nn.Mo 阅读全文
posted @ 2023-09-04 18:08 绝不原创的飞龙 阅读(160) 评论(0) 推荐(0)
摘要: ![](https://img2023.cnblogs.com/blog/1898939/202309/1898939-20230904180340788-1151936298.png) ```py class MLP(torch.nn.Module): """MLP. MLP will take 阅读全文
posted @ 2023-09-04 18:04 绝不原创的飞龙 阅读(73) 评论(0) 推荐(0)