FiD;ddlmZddlmZddlmZddlZddlZddl m Z ddl m Z ddl mZddlmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZd d Zddd Z Gd d e jBZ"y)) annotations)Path)AnyN) check_suffix)is_url)AxeleraBackend CoreMLBackendExecuTorchBackend MNNBackend NCNNBackend ONNXBackendONNXIMXBackendOpenVINOBackend PaddleBackendPyTorchBackend RKNNBackendTensorFlowBackendTensorRTBackendTorchScriptBackend TritonBackendc t|trtt|}t|tr|j Dcic]\}}t |t |}}}t|}t|j|k\rHt|d|dz dt|jdt|jdt|dt rY|djdrEddl m}m}|j!|d z d }|j Dcic] \}}||| }}}|Scc}}wcc}}w) a8Check class names and convert to dict format if needed. Args: names (list | dict): Class names as list or dict format. Returns: (dict): Class names in dict format with integer keys and string values. Raises: KeyError: If class indices are invalid for the dataset size. z(-class dataset requires class indices 0-rz%, but you have invalid class indices -z defined in your dataset YAML.rn0)ROOTYAMLzcfg/datasets/ImageNet.yamlmap) isinstancelistdict enumerateitemsintstrlenmaxkeysKeyErrormin startswithultralytics.utilsrrload)nameskvnrr names_maps \/var/www/pod-logistic/pod-ai/venv/lib/python3.12/site-packages/ultralytics/nn/autobackend.pycheck_class_namesr3"s*%Yu%&%,1KKM:DAqQQ:: J uzz|  !#=a!eWDiuzz|$%Qs5::<'8&99WY  eAh $q)<)>> model = AutoBackend(model="yolo26n.pt", device="cuda") >>> results = model(img) pt torchscriptonnxdnnopenvinoenginecoreml saved_modelpbtfliteedgetpupaddlemnnncnnimxrknntriton) executorchaxeleraz yolo26n.ptcpuFNTct |t|tjrdn|j ||}||dvz}t|t jrFt jjr(|jdk7r|dvrt jd}||d} |dk(r td||jvr dd l m} td |d | d d |dk(r || d<|| d<n |dvr|| d<|j||fi| |_|dv|_||_|j j&st)||j _t+|j j&|j _y)afInitialize the AutoBackend for inference. Args: model (str | torch.nn.Module): Path to the model weights file or a module instance. device (torch.device): Device to run the model on. dnn (bool): Use OpenCV DNN module for ONNX inference. data (str | Path, optional): Path to the additional data.yaml file containing class names. fp16 (bool): Enable half-precision inference. Supported only on specific backends. fuse (bool): Fuse Conv2D + BatchNorm layers for optimization. verbose (bool): Enable verbose logging. r@>r@rBrErPrDrArS>r@rBrErKrA)devicefp16tfjsz7Ultralytics TF.js inference is not currently supported.rexport_formatszmodel='z9' is not a supported model format. Ultralytics supports: Formatz9 See https://docs.ultralytics.com/modes/predict for help.fuseverbose>rHrCrIrJrGformat>rHrOrFrIrJrGN)super__init__rnnModule _model_typetorchrUcuda is_availabletypeNotImplementedError _BACKEND_MAPultralytics.engine.exporterrY TypeErrorbackendnhwcr]r-r=r3) selfmodelrUrCr;rVr[r\r]backend_kwargsrY __class__s r2r_zAutoBackend.__init__s, #E29954;K;KESV;W WWW vu|| , '') u$OO\\%(F%+D9 V %&_` ` ** * B%!))7)9()C(DEKL  T>%)N6 "(/N9 % H H'-N8 $0t((0I.I ZZ  ||!!!4T!:DLL .t||/A/AB cd|jvr,t|j|rt|j|St||S)awDelegate attribute access to the backend. This allows AutoBackend to transparently expose backend attributes without explicit copying. Args: name: Attribute name to look up. Returns: The attribute value from the backend. Raises: AttributeError: If the attribute is not found in backend. rk)__dict__hasattrrkgetattrr^ __getattr__)rmnamerps r2rvzAutoBackend.__getattr__s?  %'$,,*E4<<. .w"4((rqc |jr|jdddd}|jjr-|jt j k7r|j}i}|jdk(r|||d|}|jj|fi|}t|ttfrt|jdk(rg|jdk(st|dk(rJ|dj d|dj dz d z }t#|D cic]} | d |  c} |_t|dk(r|j%|dS|D cgc]} |j%| c} S|j%|Scc} wcc} w) a:Run inference on an AutoBackend model. Args: im (torch.Tensor): The image tensor to perform inference on. augment (bool): Whether to perform data augmentation during inference. visualize (bool): Whether to visualize the output predictions. embed (list, optional): A list of layer indices to return embeddings from. **kwargs (Any): Additional keyword arguments for model configuration. Returns: (torch.Tensor | list[torch.Tensor]): The raw output tensor(s) from the model. rrr@)augment visualizeembedr6segmentr7)rlpermuterkrVdtypercfloat16halfr]forwardrrtupler%r-taskshaper: from_numpy) rmimr{r|r}kwargsforward_kwargsyncr<xs r2rzAutoBackend.forwardsQ( 99Aq!Q'B <<  U]]!:B ;;$ )0ySXc\bcN DLL  6~ 6 a$ '4::#%499 +ASVq[qTZZ]QqTZZ]2Q66;Bi@a5n@ ,/FaK4??1Q4( \Z[=\UVdooa>P=\ \??1% %A=\s  FFct|tjr.tj|j |j S|S)zConvert a NumPy array to a torch tensor on the model device. Args: x (np.ndarray | torch.Tensor): Input array or tensor. Returns: (torch.Tensor): Tensor on `self.device`. )rnpndarrayrctensortorU)rmrs r2rzAutoBackend.from_numpys43=Q 2Ku||A!!$++.RQRRrqcddlm}|jdvr|jjdk7s|jdk(rt j ||jrt jnt j|jd}t|jdk(rdnd D]T}|j|t jd d d |j }|d d d dfxx|dzcc<||Vy y y )zWarm up the model by running forward pass(es) with a dummy input. Args: imgsz (tuple[int, int, int, int]): Dummy input shape in (batch, channels, height, width) format. r)non_max_suppression>rHr@rBrErPrGrArSrP)rrUrAryrT)rUNr) ultralytics.utils.nmsrr]rUrfrcemptyrVrfloatr:rrand)rmimgszrr_ warmup_boxess r2warmupzAutoBackend.warmup&s > ;;` ` KK   %)@e5:: \`\g\ghB } <1!D 2 R $zz!RDKKH QU#uRy0##L1  2*A arqcB ddlm}|d}t|st|ts t ||t |j}|Dcgc]}||vc} dxx|jdzcc< dxx d zcc<t fdt|d Dd }|d k(rd }|S|d k(r|rd}|St sHddl m }||}t|jr%t|j r|j"dvrd}|Scc}w)aTake a path to a model file and return the model format string. Args: p (str): Path to the model file. dnn (bool): Whether to use OpenCV DNN module for ONNX inference. Returns: (str): Model format string (e.g., 'pt', 'onnx', 'engine', 'triton'). Examples: >>> fmt = AutoBackend._model_type("path/to/model.onnx") >>> assert fmt == "onnx" rrXSuffixz.mlmodel c34K|]\}}|s |ywN).0r<ftypess r2 z*AutoBackend._model_type..PsYTQPUVWPXqYs ArgumentNrr@rBrC)urlsplit>grpchttprP)rirYrrr$rrrwendswithnextr!any urllib.parserboolnetlocpathscheme) prCrYsfrwsr]rurlrs @r2rbzAutoBackend._model_type8s  ?  h 'ayAs!3 B Aw||$&'qd' aDMM*-- aaL YY~/? /K%LY[_` S=F v #F U -1+CCJJDNszzEU7U! (s Dct|jdrDt|jjdr$|jjjt| S)z6Set the backend model to evaluation mode if supported.rneval)rtrkrnrr^)rmrps r2rzAutoBackend.eval]sF 4<< )gdll6H6H&.Q LL   # # %w|~rqct||}t|jdrt |jj t jrk|jj j|t|jj jj|j_ |S)aApply a function to backend.model parameters, buffers, and tensors. This method extends the functionality of the parent class's _apply method by additionally resetting the predictor and updating the device in the model's overrides. It's typically used for operations like moving the model to a different device or changing its precision. Args: fn (Callable): A function to be applied to the model's tensors. This is typically a method like to(), cpu(), cuda(), half(), or float(). Returns: (AutoBackend): The model instance with the function applied and updated attributes. rn) r^_applyrtrkrrnr`rar parametersrU)rmfnrps r2rzAutoBackend._applycsw~b! 4<< )j9K9KRYY.W LL   % %b )"&t||'9'9'D'D'F"G"N"NDLL  rq)rnzstr | torch.nn.ModulerUz torch.devicerCrr;str | Path | NonerVrr[rr\r)rwr$returnr)FFN) r torch.Tensorr{rr|rr}z list | Nonerrrz!torch.Tensor | list[torch.Tensor])rznp.ndarray | torch.Tensorrr))rrzr)rztuple[int, int, int, int]rNone)zpath/to/model.ptF)rr$rCrrr$)rr?)!__name__ __module__ __qualname____doc__rrrrrr rrr r rrrr r rhrcno_gradrUr_rvrrr staticmethodrbrr __classcell__)rps@r2r?r?Us0d n)   {  O  /  - (  # $ - z   ~  !" -#$(!'L,U]]_(4+u||E2"&?C$?C?C ?C  ?C  ?C?C?C?CB),! && &&&& &&  &&  && +&&P S2$""H rqr?)r-z list | dictrdict[int, str]r)r;rrr)# __future__rpathlibrtypingrnumpyrrctorch.nnr`r8rultralytics.utils.downloadsrbackendsr r r r r rrrrrrrrrrr3r=rar?rrqr2rsT# 1.&>0(`"))`rq