FinddlmZddlZddlZddlmZddlZddlmZddl m Z ddl m Z Gdd e Z y) ) annotationsN)Path)LOGGER)check_requirements) BaseBackendc eZdZdZddZddZy) MNNBackendzMNN (Mobile Neural Network) inference backend. Loads and runs inference with MNN models (.mnn files) using the Alibaba MNN framework. Optimized for mobile and edge deployment with configurable thread count and precision. ctjd|dtdddl}ddt j dzd zd }|j j|f}|j j|gg|d |_ |j|_ |jj}d |vr) |jtj|d yy#tj$rYywxYw)zLoad an Alibaba MNN model from a .mnn file. Args: weight (str | Path): Path to the .mnn model file. zLoading z for MNN inference...MNNrNlowCPUr) precisionbackend numThreadT)runtime_manager rearrangebizCode)rinforr os cpu_countnncreate_runtime_managerload_module_from_filenetexprget_infoapply_metadatajsonloadsJSONDecodeError)selfweightr configrtrs ]/var/www/pod-logistic/pod-ai/venv/lib/python3.12/site-packages/ultralytics/nn/backends/mnn.py load_modelzMNNBackend.load_models  hvh&;<=5!$bllnWXFX]^E^_ VV * *F9 566//BPR^b/cHH xx  "   ##DJJtI$?@ ''  s8'C!!C76C7c|jj|j|j}|jj |g}|Dcgc] }|j j"c}Scc}w)zRun inference using the MNN runtime. Args: im (torch.Tensor): Input image tensor in BCHW format, normalized to [0, 1]. Returns: (list): Model predictions as a list of numpy arrays. )rconstdata_ptrshaper onForwardreadcopy)r#im input_var output_varxs r'forwardzMNNBackend.forward/sXIIOOBKKM288< XX'' 4 )34A 444s%A=N)r$z str | PathreturnNone)r0z torch.Tensorr5list)__name__ __module__ __qualname____doc__r(r4r'r r s . 5r=r ) __future__rr rpathlibrtorchultralytics.utilsrultralytics.utils.checksrbaserr r<r=r'rDs+#  $7*5*5r=