aten::tril_ ｜ PyTorchOpMapper ｜Model is not supported yet.

[dimensionplane]

感谢您参与 X2Paddle 社区！ 问题模版为了 X2Paddle 能更好的迭代，例如新功能发布、 RoadMaps 和错误跟踪. :smile_cat:

问题描述

to be incorrect. Recommended usage would be tensor.shape[0]. Passing a tensor of different shape might lead to errors or silently give incorrect results. if len(presents) == 0:

========= 1 OPs are not supported yet =========== ========== aten::tril_ ============ Traceback (most recent call last): File "/jfs/wangxueyang/codegeex4/git_paddle_run.py", line 18, in pytorch2paddle(pt_model, save_dir , jit_type, [inputs["input_ids"], inputs["attention_mask"],inputs["position_ids"]]) File "/jfs/wangxueyang/conda/codegeex_9b_b/lib/python3.10/site-packages/x2paddle/convert.py", line 399, in pytorch2paddle mapper = PyTorchOpMapper(model) File "/jfs/wangxueyang/conda/codegeex_9b_b/lib/python3.10/site-packages/x2paddle/op_mapper/pytorch2paddle/pytorch_op_mapper.py", line 44, in init raise Exception("Model is not supported yet.") Exception: Model is not supported yet.

具体信息

• 转换模型后用处
  • [ ] 使用 Paddle 框架/ PaddleInference 推理预测
  • [ ] 使用 Paddle-Lite 做移动端推理
  • [ ] 转换预训练参数，再使用 Paddle 进行模型开发
• 模型来源 codegeex4-all-9b：https://hf-mirror.com/THUDM/codegeex4-all-9b/tree/main
• 应用场景 用于移动端业务下的检测业务
• 版本信息 PaddlePaddle => :3.0.0b1: X2Paddle => :1.6.0: 来源框架版本PyTorch=> :2.3.1:
• 您的联系方式(邮箱/微信/电话)

[megemini]

这个算子木有做映射 ～ docs/inference_model_convertor/op_list.md 这里面是目前支持的算子 ～

可以在 x2paddle/op_mapper/pytorch2paddle/aten.py 里面先尝试一下实现这个算子 ～