3ji'SrSSKrSSKrSSKrSSKrSSKJrJrJrJ r J r SSK r SSK r Sr\R "\5r/SQrS\S\4S jrS S\S \ \\ R,4S \4S jjrS \S\S\4SjrS\\\4S\\\44SjrS!S\S\S\ \\ R,4S \S\\\44 SjjrS"S\ R8R:S\S\S\ \\ R,4S\S\S\ \S \S\4SjjrS#S\\\4S\ R8R:S\S\\\44SjjrS$S\ R8R:S\S\ \ R@RBS\ \S\S \S\ \"4Sjjr#g!\a SrGNLf=f)%zhModel creation / weight loading / state_dict helpers Hacked together by / Copyright 2020 Ross Wightman N)AnyCallableDictOptionalUnionTF)clean_state_dictload_checkpointload_state_dictremap_state_dictresume_checkpointcheckpoint_pathreturnc[[RS5(dg[RR[ U55nU(aSSR U5S3$S$![ a /nN,f=f)N get_unsafe_globals_in_checkpointz Unsupported globals: z, .)hasattrtorch serializationrstr Exceptionjoin)r unsafe_globalss N/home/wildlama/miniconda3/lib/python3.13/site-packages/timm/models/_helpers.py_checkpoint_unsafe_globalsrsv 5&&(J K K,,MMcRaNbcES #DIIn$=#>a @ZXZZ s(A(( A76A7 map_location weights_onlycNU=(a [[RS5nU(aM[RR[R /5 [R "XUS9sSSS5 $[R "XUS9$!,(df  N#=f![aCnU(d[R "XS9sSnA$[S[RS35UeSnAf[Ra&nU(de[S[U5S35UeSnAff=f)N safe_globalsrr)rzAweights_only=True is not supported by this PyTorch build (torch==z). No automatic unsafe pickle fallback is performed. Upgrade PyTorch, or explicitly set weights_only=False only for trusted local checkpoints.zeweights_only=True blocked loading this checkpoint because it requires non-allowlisted pickle globals.zp No automatic unsafe pickle fallback is performed. If this checkpoint is trusted, retry with weights_only=False.) rrrrargparse Namespaceload TypeError RuntimeError __version__pickleUnpicklingErrorr)r rruse_safe_globalses r _torch_loadr+)s  $T0C0C^(T $$1183E3E2FGzz/[ghHGzz/S_``HG ::oI IOPUPaPaObch h     ! !  s)/:;!DD$textprefixcNURU5(aU[U5S$U$)N) startswithlen)r,r-s r_remove_prefixr1Hs' vCKL!! K state_dictcp0nSnUR5Hup4UHn[X55nM XAU'M U$)N)zmodule.z _orig_mod.)itemsr1)r3cleaned_state_dict to_removekvrs rrrOsJI  "Aq$A !1# r2use_emadevicecU(GaRU5(a[U5R S5(a0[ (dS5e[ RRXS9nO [XUS9nSn[U[5(aKU(aURSS5bSnO.U(aURSS5bSnOS U;aS nOS U;aS n[U(aXEOU5n[RS R!XP55 U$[R#S R!U55 [%5e) a?Load state dictionary from checkpoint file. Args: checkpoint_path: Path to checkpoint file. use_ema: Whether to use EMA weights if available. device: Device to load checkpoint to. weights_only: Whether to load only weights (torch.load parameter). Returns: State dictionary loaded from checkpoint. z .safetensorsz-`pip install safetensors` to use .safetensors)r<r rstate_dict_emaN model_emar3modelzLoaded {} from checkpoint '{}'No checkpoint found at '{}')ospathisfilerendswith_has_safetensors safetensorsr load_filer+ isinstancedictgetr_loggerinfoformaterrorFileNotFoundError)r r;r<r checkpointstate_dict_keyr3s rr r ]s"277>>/::   ( ( 8 8## T%T T#$**44_4TJ$_XdeJ j$ ' ':>>*:DAM!1Z^^K>J!,+!-J&!(%Nj&@Xbc  5<<^]^ 3::?KL!!r2r@strictremap filter_fnc@[RRU5SR5S;a.[ US5(aUR U5 g[ S5e[XX7S9nU(a [X5nOU(aU"X5nURXS9n U $)a Load checkpoint into model. Args: model: Model to load checkpoint into. checkpoint_path: Path to checkpoint file. use_ema: Whether to use EMA weights if available. device: Device to load checkpoint to. strict: Whether to strictly enforce state_dict keys match. remap: Whether to remap state dict keys by order. filter_fn: Optional function to filter state dict. weights_only: Whether to load only weights (torch.load parameter). Returns: Incompatible keys from model.load_state_dict(). )z.npzz.npyload_pretrainedz"Model cannot load numpy checkpointN)r<r)rS) rBrCsplitextlowerrrXNotImplementedErrorr r ) r@r r;r<rSrTrUrr3incompatible_keyss rr r s2 ww(,2248HH 5+ , ,  ! !/ 2 &&JK K!&dJ %j8 z1 --j-H r2 allow_reshapec 0n[UR5R5UR55HuupEupgUR5UR5:Xd(SUSURSUSURS3 5eURUR:waKU(aUR UR5nO(SUSURSUSURS3 5eXsU'M U$)a{Remap checkpoint by iterating over state dicts in order (ignoring original keys). This assumes models (and originating state dict) were created with params registered in same order. Args: state_dict: State dict to remap. model: Model whose state dict keys to use. allow_reshape: Whether to allow reshaping tensors to match. Returns: Remapped state dictionary. zTensor size mismatch z: z vs z. Remap failed.zTensor shape mismatch )zipr3r5numelshapereshape)r3r@r]out_dictkavakbvbs rr r s"H!%"2"2"4":":"N>N>PQ(2xxzRXXZ't+@BrxxjPTUWTXXZ[][c[cZdds)tt' 88rxx ZZ)j 6rd"RXXJd2$bQSQYQYPZZijju R Or2 optimizer loss_scalerlog_infocSn[RRU5(Gau[USUS9n[ U[ 5(GaSU;GaU(a[ RS5 [US5nURU5 Ub6SU;a0U(a[ RS5 URUS5 UbIURU;a9U(a[ RS5 URXsR5 S U;aGUS nS U;aUS S :aUS - nU(a'[ RS RXS 55 U$URU5 U(a$[ RS RU55 U$[ RSRU55 [5e)aResume training from checkpoint. Args: model: Model to load checkpoint into. checkpoint_path: Path to checkpoint file. optimizer: Optional optimizer to restore state. loss_scaler: Optional AMP loss scaler to restore state. log_info: Whether to log loading info. weights_only: Whether to load only weights via torch.load. Returns: Resume epoch number if available, else None. Ncpur r3z(Restoring model state from checkpoint...rhz,Restoring optimizer state from checkpoint...z2Restoring AMP loss scaler state from checkpoint...epochversionz!Loaded checkpoint '{}' (epoch {})zLoaded checkpoint '{}'rA)rBrCrDr+rIrJrLrMrr rRrNrOrP) r@r rhrirjr resume_epochrQr3s rr r s*L ww~~o&& uS_` j$ ' 'LJ,F GH)*\*BCJ  ! !* -$ )BLL!OP))*[*AB&;+E+E+SLL!UV++J7Q7Q,RS*$)'2  *z)/Dq/H A%LLL!D!K!KOho]p!qr   ! !* - 5<<_MN 3::?KL!!r2)rlT)TrlT)TrlTFNT)T)NNTT)$__doc__r!loggingrBr'typingrrrrrrsafetensors.torchrGrF ImportError getLogger__name__rL__all__rrr<boolr+r1rr nnModuler r optim Optimizerintr r2rrs 77    H % [[[27!C-.>cc c3h DcN  +0! ("("("c5<<'((" ("  #s(^ ("\+0(,!'xx'''c5<<'( '  '  'H%'' 'Z#cNxx #s(^ @6:%)! 6"xx6"6"EKK1126"c] 6"  6"  6"c]6"}sE--E98E9