在使用jit导出训练好的RNN-T模型时，报错Unknown type name 'dict'

[DaobinZhu]

命令为：
python ./wenet/bin/export_jit.py
--config /home/lsj/zdb/biye/wenet/examples/aishell/rnnt/exp/baseline/train.yaml
--checkpoint /home/lsj/zdb/biye/wenet/examples/aishell/rnnt/exp/baseline/avg_4.pt
--output_file /home/lsj/zdb/biye/wenet/examples/aishell/rnnt/exp/baseline/final.zip
--output_quant_file /home/lsj/zdb/biye/wenet/examples/aishell/rnnt/exp/baseline/final_quant.zip

报错结果为：
/home/lsj/.conda/envs/rnnt/lib/python3.8/site-packages/torch/jit/_check.py:178: UserWarning: The TorchScript type system doesn't support instance-level annotations on empty non-base types in __init__. Instead, either 1) use a type annotation in the class body, or 2) wrap the type in torch.jit.Attribute.
warnings.warn(
Traceback (most recent call last):
File "./wenet/bin/export_jit.py", line 70, in
main()
File "./wenet/bin/export_jit.py", line 54, in main
script_model = torch.jit.script(model)
File "/home/lsj/.conda/envs/rnnt/lib/python3.8/site-packages/torch/jit/_script.py", line 1324, in script
return torch.jit._recursive.create_script_module(
File "/home/lsj/.conda/envs/rnnt/lib/python3.8/site-packages/torch/jit/_recursive.py", line 559, in create_script_module
return create_script_module_impl(nn_module, concrete_type, stubs_fn)
File "/home/lsj/.conda/envs/rnnt/lib/python3.8/site-packages/torch/jit/_recursive.py", line 636, in create_script_module_impl
create_methods_and_properties_from_stubs(
File "/home/lsj/.conda/envs/rnnt/lib/python3.8/site-packages/torch/jit/_recursive.py", line 469, in create_methods_and_properties_from_stubs
concrete_type._create_methods_and_properties(
RuntimeError:
Unknown type name 'dict':
File "/home/lsj/zdb/biye/wenet/wenet/transducer/transducer.py", line 100
def forward(
self,
batch: dict,
~~~~ <--- HERE
device: torch.device,
) -> Dict[str, Optional[torch.Tensor]]:

[Mddct]

rnnt 还不支持jit导出 做推理

[DaobinZhu]

周佬，近期有做rnnt解码的打算吗

…

---原始邮件--- 发件人: "Dinghao ***@***.***&gt; 发送时间: 2024年2月24日(周六) 上午8:07 收件人: ***@***.***&gt;; 抄送: "Daobin ***@***.******@***.***&gt;; 主题: Re: [wenet-e2e/wenet] 在使用jit导出训练好的RNN-T模型时，报错Unknown type name 'dict' (Issue #2369) rnnt 还不支持jit导出 做推理 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;

[Mddct]

有的 直接是两件事情 一个是迁移w2vbert2.0 的参数，
另外一件事是把llm 的一些设计迁移过来

rnnt优先放这这两个后边

[DaobinZhu]

好的，谢谢

…

---原始邮件--- 发件人: "Dinghao ***@***.***&gt; 发送时间: 2024年2月24日(周六) 上午8:19 收件人: ***@***.***&gt;; 抄送: "Daobin ***@***.******@***.***&gt;; 主题: Re: [wenet-e2e/wenet] 在使用jit导出训练好的RNN-T模型时，报错Unknown type name 'dict' (Issue #2369) 有的 直接是两件事情 一个是迁移w2vbert2.0 的参数， 另外一件事是把llm 的一些设计迁移过来 rnnt优先放这这两个后边 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;