/jddlmZddlmZddlmZddlZddlmZddl m Z m Z m Z ddl mZGdd eZGd d eZdS) ) annotations)Path)AnyN) IS_JETSONLOGGER is_jetson) BaseBackendc@eZdZdZ ddfd ZddZ dd dZxZS)!PyTorchBackendzPyTorch inference backend for native model execution. Loads and runs inference with native PyTorch models (.pt checkpoint files) or pre-loaded nn.Module instances. Supports model layer fusion, FP16 precision, and NVIDIA Jetson compatibility. FTweightstr | Path | nn.Moduledevice torch.devicefp16boolfuseverbosecj||_||_t|||dS)aInitialize the PyTorch backend. Args: weight (str | Path | nn.Module): Path to the .pt model file or a pre-loaded nn.Module instance. device (torch.device): Device to run inference on (e.g., 'cpu', 'cuda:0'). fp16 (bool): Whether to use FP16 half-precision inference. fuse (bool): Whether to fuse Conv2D + BatchNorm layers for optimization. verbose (bool): Whether to print verbose model loading messages. N)rrsuper__init__)selfr rrrr __class__s d/home/longshao/multi-rider-rag/.venv/lib/python3.11/site-packages/ultralytics/nn/backends/pytorch.pyrzPyTorchBackend.__init__s5"   .....str | torch.nn.ModulereturnNonecddlm}t|tjjr~|jr\t|drLtr*tdr| |j }||j }| |j }n|||j |j\}}t|dr |j |_ t|d r4tt|jd nd |_t|d r |jjnt'|d i|_t|d r|jddnd|_|jr|n||D] }d|_ ||_t'|dd|_dS)zLoad a PyTorch model from a checkpoint file or nn.Module instance. Args: weight (str | torch.nn.Module): Path to the .pt checkpoint or a pre-loaded module. r)load_checkpointr)jetpack)r)rr kpt_shapestride modulenamesyamlchannelsFend2endN)ultralytics.nn.tasksr isinstancetorchnnModulerhasattrrrtorrr#maxintr$r&r'getattrr(getr)rhalffloat parameters requires_gradmodelr+)rr r r;_ps r load_modelzPyTorchBackend.load_model,s 988888 feho . . Sy ;WVV44 ;41!5!5!54#YYt{33FT\::IIdk**EE&vdk RRRHE1 5+ & & -"_DN:A%:R:RZc#el..00112666XZ +25(+C+CdU\''QVX_acIdId 9@9O9OV z1555UV  4 u{{}}!!## $ $A#AOO ui77 rNim torch.Tensoraugment visualizeembed list | Nonekwargsr!torch.Tensor | list[torch.Tensor]c &|j|f|||d|S)ayRun native PyTorch inference with support for augmentation, visualization, and embeddings. Args: im (torch.Tensor): Input image tensor in BCHW format, normalized to [0, 1]. augment (bool): Whether to apply test-time augmentation. visualize (bool): Whether to visualize intermediate feature maps. embed (list | None): List of layer indices to extract embeddings from, or None. **kwargs (Any): Additional keyword arguments passed to the model forward method. Returns: (torch.Tensor | list[torch.Tensor]): Model predictions as tensor(s). )rArBrCr;)rr?rArBrCrEs rforwardzPyTorchBackend.forwardKs(tz"Zg%ZZSYZZZr)FTT) r rrrrrrrrr)r rrr)FFN) r?r@rArrBrrCrDrErrrF__name__ __module__ __qualname____doc__rr>rI __classcell__rs@rr r s ///////*8888@fj[[[[[[[[[rr c6eZdZdZddfd Zdd ZddZxZS)TorchScriptBackendaPyTorch TorchScript inference backend for serialized model execution. Loads and runs inference with TorchScript models (.torchscript files) created via torch.jit.trace or torch.jit.script. Supports FP16 precision and embedded metadata extraction. Fr str | PathrrrrcNt|||dS)aInitialize the TorchScript backend. Args: weight (str | Path): Path to the .torchscript model file. device (torch.device): Device to run inference on (e.g., 'cpu', 'cuda:0'). fp16 (bool): Whether to use FP16 half-precision inference. N)rr)rr rrrs rrzTorchScriptBackend.__init__ds' .....rstrrrcddl}ddl}tjd|dddi}tj|||j|_|j r|j n|j |dr3| | |dd dSdS) zLoad a TorchScript model from a .torchscript file with optional embedded metadata. Args: weight (str): Path to the .torchscript model file. rNzLoading z for TorchScript inference...z config.txt) _extra_files map_locationcDt|S)N)dictitems)xs rz/TorchScriptBackend.load_model..~s\`abahahajaj\k\kr) object_hook)json torchvisionrinfor.jitloadrr;rr7r8apply_metadataloads)rr r`ra extra_filess rr>zTorchScriptBackend.load_modelns   DvDDDEEE#R( Y^^FSWS^^__ !Y> DJ,<,<,>,> | $ n    ;|+DRkRk l l m m m m m n nrr?r@rFc,||S)zRun TorchScript inference. Args: im (torch.Tensor): Input image tensor in BCHW format, normalized to [0, 1]. Returns: (torch.Tensor | list[torch.Tensor]): Model predictions as tensor(s). rH)rr?s rrIzTorchScriptBackend.forwardszz"~~r)F)r rSrrrr)r rUrr)r?r@rrFrJrPs@rrRrR]sx ///////nnnn$        rrR) __future__rpathlibrtypingrr.torch.nnr/ultralytics.utilsrrrbaser r rRrrrps#""""" ::::::::::J[J[J[J[J[[J[J[J[Z,,,,,,,,,,r