在提交新问题之前，请确保您已经在社区中搜索过相关问题，并使用了社区中提供的资源/工具后，仍未找到满意的解决方式。

环境信息

例如：
- 操作系统
- 昇腾硬件信息
- CANN软件版本
- 安装的对应软件版本

软件版本参考链接：https://gitcode.com/Ascend/RecSDK/blob/develop_torch_benchmark/torch2.6.0_examples_benchmark/develop/README.md

[root@liteserver-c2d2 gr_nv]# pip list | grep torch
gpytorch 1.15.2
hybrid-torchrec 1.1.0
mindspeed 0.8.0 .....RecSDK/torch2.6.0_examples_benchmark/develop/gr_nv/MindSpeed
torch 2.6.0+cpu
torch_npu 2.6.0
torchmetrics 1.0.3
torchrec 1.1.0+npu
torchrec-embcache 1.1.0
torchx 0.7.0

[root@liteserver-c2d2 gr_nv]# pip list | grep trans
npu_ops_transformer 1.0.0
transformers 5.11.0

您是如何安装操作的

# 示例：
pip install ... # 以及其他完整的安装命令

参考https://gitcode.com/Ascend/RecSDK/blob/develop_torch_benchmark/torch2.6.0_examples_benchmark/develop/README.md执行bash run.sh

报错日志：
[rank3]: File "/...../RecSDK/torch2.6.0_examples_benchmark/develop/gr_nv/recsys-examples/examples/hstu/model/ranking_gr.py", line 170, in get_logit_and_labels
[rank3]: embeddings = self._embedding_collection(batch.features)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
[rank3]: return self._call_impl(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl
[rank3]: return forward_call(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torchrec_embcache/distributed/embedding.py", line 982, in forward
[rank3]: return self.compute_and_output_dist(ctx, post_dist_input)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torchrec_embcache/distributed/embedding.py", line 595, in compute_and_output_dist
[rank3]: lookup_ret = lookup(features)
[rank3]: ^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
[rank3]: return self._call_impl(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl
[rank3]: return forward_call(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torchrec/distributed/embedding_lookup.py", line 285, in forward
[rank3]: embeddings.append(emb_op(features).view(-1))
[rank3]: ^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
[rank3]: return self._call_impl(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl
[rank3]: return forward_call(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/hybrid_torchrec/distributed/batched_embedding_kernel.py", line 891, in forward
[rank3]: return self.emb_module(
[rank3]: ^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1739, in _wrapped_call_impl
[rank3]: return self._call_impl(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1750, in _call_impl
[rank3]: return forward_call(*args, **kwargs)
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/hybrid_torchrec/distributed/batched_embedding_kernel.py", line 393, in forward
[rank3]: result = invokers.lookup_adam.invoke(
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/hybrid_torchrec/hybrid_lookup_invoke/hybrid_lookup_adam.py", line 29, in invoke
[rank3]: return torch.ops.fbgemm.split_embedding_codegen_lookup_adam_function(
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: File "/usr/local/python3.11.0/lib/python3.11/site-packages/torch/_ops.py", line 1123, in call
[rank3]: return self._op(*args, **(kwargs or {}))
[rank3]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank3]: RuntimeError: call aclnnSplitEmbeddingCodegenForwardUnweighted failed, detail:[PID: 138879] 2026-06-12-11:48:21.466.442 AclNN_Parameter_Error(EZ1001): Get regInfo failed, The binary_info_config.json of socVersion [ascend910_93] does not support opType [SplitEmbeddingCodegenForwardUnweighted].
[rank3]: TraceBack (most recent call last):
[rank3]: Check nnopExecutor != nullptr failed

欢迎加入社区，感谢您对社区的贡献 🎉!