wiNddlZddlZddlZddlZddlZddlmZddlmZddl m Z ddl m Z m Z mZmZmZmZddlZddlZddlmZmZddlmZddlmZmZd d lmZd d lmZmZm Z m!Z!m"Z"m#Z#m$Z$m%Z%m&Z&m'Z'm(Z(m)Z)m*Z*d d l+m,Z,m-Z-m.Z.d dl/m0Z0m1Z1m2Z2m3Z3m4Z4e*j5e6Z7ej8dZ9e)ddrdZ:ndZ:e(rddl;Z;dejj<dej=fdZ>dejj<dej?fdZ@Gddejj<e,ZAGddeAZBdS)N) OrderedDict)partial)Path)AnyCallableListOptionalTupleUnion) create_repo"split_torch_state_dict_into_shards)validate_hf_hub_args)Tensornn) __version__) CONFIG_NAMEFLAX_WEIGHTS_NAMESAFE_WEIGHTS_INDEX_NAMESAFETENSORS_WEIGHTS_NAMEWEIGHTS_INDEX_NAME WEIGHTS_NAME _add_variant_get_checkpoint_shard_files_get_model_file deprecateis_accelerate_availableis_torch_versionlogging)PushToHubMixinload_or_create_model_cardpopulate_model_card)_determine_device_map_fetch_index_file_load_state_dict_into_modelload_model_dict_into_metaload_state_dictz(.*?)-\d{5}-of-\d{5}>=1.9.0TF parameterreturnc tj||}t |jS#t $rndtjj dttttffd}||}t |}|djcYSwxYw)Nmoduler,cLd|jD}|S)NcDg|]\}}tj|||fStorch is_tensor.0kvs o/root/.openclaw/workspace/chatterbox_venv_py311/lib/python3.11/site-packages/diffusers/models/modeling_utils.py zHget_parameter_device..find_tensor_attributes..V0WWWAEOTUDVDVWq!fWWW__dict__itemsr.tupless r9find_tensor_attributesz4get_parameter_device..find_tensor_attributesU(WW)>)>)@)@WWWFMr<get_members_fnr#) itertoolschain parametersbuffersnextdevice StopIterationr3rModulerr strr_named_members)r+parameters_and_buffersrBgen first_tuples r9get_parameter_devicerSNs %!*1E1E1G1GIZIZI\I\!]!]*++22  % % % 58? tE#v+DV?W    &&6L&MM3ii 1~$$$$ %sA AA5CCc t|}t|dkr |djSt|}t|dkr |djSdS#t $rndt jjdttttffd}| |}t|}|djcYSwxYw)Nrr.r,cLd|jD}|S)NcDg|]\}}tj|||fSr1r2r5s r9r:zGget_parameter_dtype..find_tensor_attributes..lr;r<r=r@s r9rBz3get_parameter_dtype..find_tensor_attributeskrCr<rDr#)tuplerHlendtyperIrLr3rrMrr rNrrOrJ)r+paramsrIrBrQrRs r9get_parameter_dtyper[^s$y++--.. v;;??!9? " ))++,, w<=C>ceZdZdZeZgdZdZdZdZ fdZ de de ffd Z edefd Zd.d Zd.d Zd eddfd Zd.dZd.dZ d/d edeeddfdZd/deeddfdZd.dZ d0dee ejfdedeededee deee fdefdZe e!deee ejffdZ"e d1d e#dee ejfd!efd"Z$e d#Z%d$e fd%Z&ede'j(fd&Z(ede'j)fd'Z)d2d(ed)edefd*Z*d e#ddfd+Z+d.d,Z,d.d-Z-xZ.S)3 ModelMixina$ Base class for all models. [`ModelMixin`] takes care of storing the model configuration and provides methods for loading, downloading and saving models. - **config_name** ([`str`]) -- Filename to save a model to when calling [`~models.ModelMixin.save_pretrained`]. )_diffusers_version _class_name _name_or_pathFNcHtdSN)super__init__)self __class__s r9rdzModelMixin.__init__s r<namer,c Vd|jvot|jd|}||jv}|rY|sWd|dt|jd|dt|jd|d }t dd |d d |j|St |S) a~The only reason we overwrite `getattr` here is to gracefully deprecate accessing config attributes directly. See https://github.com/huggingface/diffusers/pull/3129 We need to overwrite __getattr__ here in addition so that we don't trigger `torch.nn.Module`'s __getattr__': https://pytorch.org/docs/stable/_modules/torch/nn/modules/module.html#Module _internal_dictzAccessing config attribute `z` directly via 'z1' object attribute is deprecated. Please access 'z' over 'z,'s config object instead, e.g. 'unet.config.z'.zdirect config name accessz1.0.0F) standard_warn stacklevel)r>hasattrtype__name__rrirc __getattr__)rerg is_in_config is_attributedeprecation_messagerfs r9rpzModelMixin.__getattr__s(4=8kWT]ScEdfj=k=k t},  -  -#u#u#uW[\`WaWaWj#u#u^b#u#ulpqulvlvl#u#umq#u#u#u  17z7ModelMixin.is_gradient_checkpointing..s6mmYZ71677TA.fn_recursive_set_npu_flash_attentionsav<== :225999** < <44U;;;; < .fn_recursive_set_mem_effsdvLMM XBB5,WWW** 0 0((//// 0 0r<r)rerrr.rs `` @r9rz6ModelMixin.set_use_memory_efficient_attention_xformerss  0UX_ 0 0 0 0 0 0 0 0mmoo 1 1F&%(/22 1((000 1 1r<c2|d|dS)uE Enable memory efficient attention from [xFormers](https://facebookresearch.github.io/xformers/). When this option is enabled, you should observe lower GPU memory usage and a potential speed up during inference. Speed up during training is not guaranteed. ⚠️ When memory efficient attention and sliced attention are both enabled, memory efficient attention takes precedent. Parameters: attention_op (`Callable`, *optional*): Override the default `None` operator for use as `op` argument to the [`memory_efficient_attention()`](https://facebookresearch.github.io/xformers/components/ops.html#xformers.ops.memory_efficient_attention) function of xFormers. Examples: ```py >>> import torch >>> from diffusers import UNet2DConditionModel >>> from xformers.ops import MemoryEfficientAttentionFlashAttentionOp >>> model = UNet2DConditionModel.from_pretrained( ... "stabilityai/stable-diffusion-2-1", subfolder="unet", torch_dtype=torch.float16 ... ) >>> model = model.to("cuda") >>> model.enable_xformers_memory_efficient_attention(attention_op=MemoryEfficientAttentionFlashAttentionOp) ``` TNr)rers r9*enable_xformers_memory_efficient_attentionz5ModelMixin.enable_xformers_memory_efficient_attentions!D 88|LLLLLr<c0|ddS)zs Disable memory efficient attention from [xFormers](https://facebookresearch.github.io/xformers/). FNrr{s r9+disable_xformers_memory_efficient_attentionz6ModelMixin.disable_xformers_memory_efficient_attentions 88?????r<T10GBsave_directoryis_main_process save_functionsafe_serializationvariantmax_shard_size push_to_hubc  tj|r td|ddS|rt nt } t| |} | d} t| dvr*| ddzd | ddz} ntd | d tj |d |r| d d} | dd} | dd}| dd}| d|tjjd}t|d | |j}|}|r||| t' || }|r"tj|D] }||jvrtj ||}tj|s_| dddd}|dd}|dddd}||r.t2|tj||jD]o\}} fd|D}tj ||}|r%t:j||ddiZt=j ||p|j!r|j"|j#d}|rtHntJ}tj |t||}tM|dd !5}tOj(|d"d #d$z}|)|dddn #1swxYwYt*d%|d&t|jd'|dn=tj || }t*d(||rqtW||)}tY|}| t[|d*.|/|||| |+dSdS),a Save a model and its configuration file to a directory so that it can be reloaded using the [`~models.ModelMixin.from_pretrained`] class method. Arguments: save_directory (`str` or `os.PathLike`): Directory to save a model and its configuration file to. Will be created if it doesn't exist. is_main_process (`bool`, *optional*, defaults to `True`): Whether the process calling this is the main process or not. Useful during distributed training and you need to call this function on all processes. In this case, set `is_main_process=True` only on the main process to avoid race conditions. save_function (`Callable`): The function to use to save the state dictionary. Useful during distributed training when you need to replace `torch.save` with another method. Can be configured with the environment variable `DIFFUSERS_SAVE_MODE`. safe_serialization (`bool`, *optional*, defaults to `True`): Whether to save the model using `safetensors` or the traditional PyTorch way with `pickle`. variant (`str`, *optional*): If specified, weights are saved in the format `pytorch_model..bin`. max_shard_size (`int` or `str`, defaults to `"10GB"`): The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size lower than this size. If expressed as a string, needs to be digits followed by a unit (like `"5GB"`). If expressed as an integer, the unit is bytes. Note that this limit will be decreased after a certain period of time (starting from Oct 2024) to allow users to upgrade to the latest version of `diffusers`. This is to establish a common default size for this argument across different libraries in the Hugging Face ecosystem (`transformers`, and `accelerate`, for example). push_to_hub (`bool`, *optional*, defaults to `False`): Whether or not to push your model to the Hugging Face Hub after saving it. You can specify the repository you want to push to with `repo_id` (will default to the name of `save_directory` in your namespace). kwargs (`Dict[str, Any]`, *optional*): Additional keyword arguments passed along to the [`~utils.PushToHubMixin.push_to_hub`] method. zProvided path (z#) should be a directory, not a fileN.)rrjrz {suffix}.r#zInvalid z provided.T)exist_okcommit_messageprivateF create_prtokenrepo_id)rrr)rfilename_patternz.binz .safetensorsz{suffix}c"i|] }|| Sr1r1)r6tensor state_dicts r9 z.ModelMixin.save_pretrained..os FFFFVZ/FFFr<formatpt)metadata)r weight_mapwzutf-8)encodingr)indent sort_keys z:The model is bigger than the maximum size per checkpoint (z) and is going to be split in z^ checkpoint shards. You can find where each parameters has been saved in the index located at zModel weights saved in )rz README.md)rrr)0ospathisfileloggererrorrrrsplitrXjoinrmakedirspopsepr r save_configrr listdirfilename_to_tensorskeysreplace startswith _REGEX_SHARD fullmatchremover? safetensorsr3 save_filesave is_shardedrtensor_to_filenamerropenjsondumpswriteinfor!r"ras_posix_upload_folder)!rerrrrrrrkwargs weights_nameweight_name_splitweights_name_patternrrrrr model_to_savestate_dict_splitfilename full_filenameweights_without_extfilename_without_exttensorsshardfilepathindexsave_index_filefcontentpath_to_weights model_cardrs! @r9save_pretrainedzModelMixin.save_pretrained sX 7>>. ) )  LL^>^^^ _ _ _ F3EW//< #L':: (..s33  ! !V + +#4Q#7+#EQbcdceceQfHgHg#g @ @@@AA A NT2222  `#ZZ(8$??NjjE22G ;66IJJw--EjjN,@,@,M,Mb,QRRG!'D'QVWWW_G   6  % %n 5 5 5#--// > ~H\     -J~66 - -/CHHJJJJ " ^X F F w~~m44&:&B&B62&N&N&V&VWegi&j&j#&9&A&A*b&Q&Q#'/'7'7'C'C'K'KN\^'_'_$''(;<<-$../CDDPIm,,,!1!E!K!K!M!M , , HgFFFFgFFFEw||NH==H! ,!++E8xQUFV+WWWW 5(++++  & E,5.AE:Lc55QcO gll><Y`;a;abbOosW=== !*U1EEEL    ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! KK7^77 0 DEE77$3777     !gll><HHO KKC/CC D D D  27%HHHJ,Z88J OOD==FFHH I I I   -#        s0PP Ppretrained_model_name_or_pathc J,|dd}|dd}|dd}|dd}|dd}|dd}|d d} |d d} |d d} |d d} |d d} |dd}|dd}|dd}|dd}|dd}|dt}|dd}|dd}d}|d}d}|r*tsd}td|tst d|t ddst d|durt ddst d|dur|td|dt|tj rd|i}nt|tr=|d vr9 dtj |i}nO#t$rtd!|d"wxYwt|tr|d#krtd$d|i}||d}n|std%|r!|t dd&std'|}td(d)d*}|j|f|dd|||| | | ||d+ |\}}}d}d}t j|}t'|||pd||||||| | | ||,}||rd}|r|rtd-d}|rBt+|t,||||| | | |||. }|j|fi|} d/d0lm}!|!| |} n|rt5|||||| | || |pd1 \}"}#n|r|s t+|t7t8|||||| | | |||. }nT#t:$rG}$td2|d3|$|std4Yd}$~$nd}$~$wwxYw|0|s.t+|t7t>|||||| | | |||. }|rTtAj!5|j|fi|} dddn #1swxYwY|?|s|j+d?d:(|(gntY| ||| } tAj-| |s|n|"||||| dd@ n#t\$r}$dAt|$vrrtdBt|$dC|dD| /tAj-| |||||| E| 0n|$Yd}$~$nd}$~$wwxYwggggdF})n[|j|fi|} tE||6}&| #|&|1| |&|||G\} }'}(}*}+|'|(|*|+dF})| = 1.9.0 else `False`): Speed up model loading only loading the pretrained weights and not initializing the weights. This also tries to not use more than 1x model size in CPU memory (including peak memory) while loading the model. Only supported for PyTorch >= 1.9.0. If you are using an older version of PyTorch, setting this argument to `True` will raise an error. variant (`str`, *optional*): Load weights from a specified `variant` filename such as `"fp16"` or `"ema"`. This is ignored when loading `from_flax`. use_safetensors (`bool`, *optional*, defaults to `None`): If set to `None`, the `safetensors` weights are downloaded if they're available **and** if the `safetensors` library is installed. If set to `True`, the model is forcibly loaded from `safetensors` weights. If set to `False`, `safetensors` weights are not loaded. To use private or [gated models](https://huggingface.co/docs/hub/models-gated#gated-models), log-in with `huggingface-cli login`. You can also activate the special ["offline-mode"](https://huggingface.co/diffusers/installation.html#offline-mode) to use this method in a firewalled environment. Example: ```py from diffusers import UNet2DConditionModel unet = UNet2DConditionModel.from_pretrained("runwayml/stable-diffusion-v1-5", subfolder="unet") ``` If you get the error message below, you need to finetune the weights for your downstream task: ```bash Some weights of UNet2DConditionModel were not initialized from the model checkpoint at runwayml/stable-diffusion-v1-5 and are newly initialized because the shapes did not match: - conv_in.weight: found shape torch.Size([320, 4, 3, 3]) in the checkpoint and torch.Size([320, 9, 3, 3]) in the model instantiated You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference. ``` cache_dirNignore_mismatched_sizesFforce_download from_flaxresume_downloadproxiesoutput_loading_infolocal_files_onlyrrevision torch_dtype subfolder device_map max_memoryoffload_folderoffload_state_dictlow_cpu_mem_usageruse_safetensorsTa,Cannot initialize model with low cpu memory usage because `accelerate` was not found in the environment. Defaulting to `low_cpu_mem_usage=False`. It is strongly recommended to install `accelerate` for faster and less memory-intense model loading. You can do so with: ``` pip install accelerate ``` .zLoading and dispatching requires `accelerate`. Please make sure to install accelerate or set `device_map=None`. You can install accelerate with `pip install accelerate`.r)r*ztLoading and dispatching requires torch >= 1.9.0. Please either update your PyTorch version or set `device_map=None`.z~Low memory initialization requires torch >= 1.9.0. Please either update your PyTorch version or set `low_cpu_mem_usage=False`.zEYou cannot set `low_cpu_mem_usage` to `False` while using device_map=zO for loading and dispatching. Please make sure to set `low_cpu_mem_usage=True`.r)autobalancedbalanced_low_0 sequentialzWhen passing device_map as a string, the value needs to be a device name (e.g. cpu, cuda:0) or 'auto', 'balanced', 'balanced_low_0', 'sequential' but found rrznYou can't pass device_map as a negative int. If you want to put the model on the cpu, pass device_map = 'cpu' z>Passing along a `device_map` requires `low_cpu_mem_usage=True`z1.10z=`low_cpu_mem_usage` and `device_map` require PyTorch >= 1.10.modelpytorch diffusers file_type framework rreturn_unused_kwargsreturn_commit_hashrrrrrrr user_agent)is_localrrrrrrrrrrrr commit_hashzFLoading of sharded checkpoints is not supported when `from_flax=True`.) rrrrrrrrrrrr#)%load_flax_checkpoint_in_pytorch_model)rrrrrrrrz(An error occurred while trying to fetch : zXDefaulting to unsafe serialization. Pass `allow_pickle=False` to raise an error instead.cpu)rz Cannot load z from z+ because the following keys are missing: z, z. Please make sure to pass `low_cpu_mem_usage=False` and `device_map=None` if you want to randomly initialize those weights or else make sure your checkpoint file is correct.)rKrYmodel_name_or_pathc>g|]}tj||Srb)research)r6r7pats r9r:z.ModelMixin.from_pretrained..s,.g.g.gQRYWZ\]M^M^MfqMfMfMfr<zESome weights of the model checkpoint were not used when initializing z: )rrrrY force_hooksstrictz#'Attention' object has no attributezTaking `z@` while using `accelerate.load_checkpoint_and_dispatch` to mean a was saved with deprecated attention block weight names. We will load it with the deprecated attention block names and convert them on the fly to the new attention block format. Please re-save the model after this conversion, so we don't have to do the on the fly renaming in the future. If the model is from a hub checkpoint, please also re-upload it or open a PR on the original repository.)rrrrY) missing_keysunexpected_keysmismatched_keys error_msgs)rzA needs to be of type `torch.dtype`, e.g. `torch.float16`, but is )r`)7r_LOW_CPU_MEM_USAGE_DEFAULTrrwarningNotImplementedErrorrrrr3rKrN RuntimeErrorintr load_configrrisdirr%is_filerr from_configmodeling_pytorch_flax_utilsrrrrIOErrorrr accelerateinit_empty_weightsr($_convert_deprecated_attention_blockssetrrrXrr'"_keys_to_ignore_on_load_unexpectedror$load_checkpoint_and_dispatchAttributeError1_temp_convert_self_to_deprecated_attention_blocks6_undo_temp_convert_self_to_deprecated_attention_blocks_load_pretrained_modelrYrntoregister_to_configeval)-clsrrrrrrrrrrrrrrrrrrrrr allow_pickle config_pathrconfig unused_kwargsrr index_filer model_filer rsharded_ckpt_cached_foldersharded_metadatae param_devicerrr  loading_infor!r"rs- @r9from_pretrainedzModelMixin.from_pretraineds VJJ{D11 "(**-F"N"N$4e<<JJ{E22  **%6==**Y--$jj)>FF!::&8$?? 7D))::j$//jj55 JJ{D11 ZZ d33 ZZ d33 $4d;;#ZZ(%>  %  NN.     !*A*C*C !%`   !*:4*I*I !%&   $ $-=dG-L-L $%.   % %**@RXbRRR  j%, / / .j)JJ  C ( ( .Z?s-s-s  %,z":":;     bT^bbb   C ( ( .A~~ E!*-  ! ($(!!& c !abbb  b%.>tV.L.L% !`aaa4 % "  .=S_ . !%#)+-!. . . . * {"  7==!>??&*Go2+)+-!#    !j&8&8&:&: !J  g) geff f { (-.#- /!1!#%'   J$COF<>zJJJ#&u'7'7'9'9'>'>'@'@#A#AC HYHYDZDZ#ZL<((1,,(`3``6S``,0IIl,C,C```'@"+)+H '''O=I#&#IhhC.g.g.g.g/.g.g.gOO?++a//VdgdpVVx|yByBCRySySxTVV"7uj*Va!b!bJ*$"?!.8XJJ>X&'1+9/A"-(,#'     *$$$ACFFJJ"NN!e3q66!e!etQ!e!e!e"SSUUU&C % * *+5/=3E&1"XXZZZZ"#G[ZZZZ9$B%'')')"$    (@@-@@,ZIII :::FFFTWTnTn1,C UoUUQ|_oz%1'6'6",      ":k5;+O+O "uuaefqararuuu  $HH[))E   /L MMM   ',& & sOJ))K7.Q&& R70=R22R7TT!$T!=!Z \8)B\33\8rrc <|}t|}t|}|} tt|t|z } tt|t|z } |} d} || ||| |}t | |}t |dkr=d|}d|vr|dz }td|jj d|t | dkrIt d|d |jj d | d |jj d |jj d n(t d|jj dt | dkr/t d|jj d|d| dnKt |dkr8t d|jj d|d|jj dt |dkrMdd|D}t d|jj d|d|d|| | ||fS)Ncg}|rX|D]U}|}||vrM||j||jkr1||||j||jf||=V|Srb)shapeappend)rmodel_state_dict loaded_keysrr!checkpoint_key model_keys r9_find_mismatched_keysz@ModelMixin._load_pretrained_model.._find_mismatched_keysxs !O& 7&1 7 7N .I"%555&~6<@PQZ@[@aaa'..+Z-G-MO_`iOjOpq'~6" "r<rz z size mismatchz_ You may consider adding `ignore_mismatched_sizes=True` in the model `from_pretrained` method.z#Error(s) in loading state_dict for z: z(Some weights of the model checkpoint at z! were not used when initializing rz, - This IS expected if you are initializing z from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model). - This IS NOT expected if you are initializing z from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).z9All model checkpoint weights were used when initializing z. zSome weights of z3 were not initialized from the model checkpoint at z and are newly initialized: zo You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.zAll the weights of z/ were initialized from the model checkpoint at zf. If your task is similar to the task the model of the checkpoint was trained on, you can already use z* for predictions without further training.rc .g|]\}}}d|d|d|dS)z- z: found shape z in the checkpoint and z in the model instantiatedr1)r6keyshape1shape2s r9r:z5ModelMixin._load_pretrained_model..sJ+VVnmmFmm6mmmr<z= and are newly initialized because the shapes did not match: ) rlistrr1r&rXrr&rfrorr$r)r;r rresolved_archive_filerrrLrM expected_keysoriginal_loaded_keysrr  model_to_loadrPr!r" error_msgmismatched_warnings r9r7z!ModelMixin._load_pretrained_modelasm!++--:??,,-- -224455 *C ..[1A1AABB s;//#m2D2DDEE  # # #*  !33 $' O 5]JOOJ z??Q   J//I)++w oU_E]oodmoopp p   ! # # NN9;X99!&!999=L99!&!999 O, 999     KKqTYTcTlqqq r r r |  q NNn5?#;nn1nnO[nnn      ! !Q & & KK-eo&>--1--CH?C[---      ! # #!%/>""  NNA5?#;AA1AA.AAA   lO_jPPr<c"tj|jj}d|D}t d|D}t |dhz }||fS)NcDi|]\}}|jtjk||Sr1defaultinspect_emptyr5s r9rz2ModelMixin._get_signature_keys..s.bbb1aiSZSaFaFaq!FaFaFar<cBh|]\}}|jtjk|Sr1r^r5s r9 z1ModelMixin._get_signature_keys..s,"c"c"cAqyT[TbGbGb1GbGbGbr<re)r` signaturerdrHr?r1r)r;objrHrequired_parametersoptional_parametersexpected_moduless r9_get_signature_keyszModelMixin._get_signature_keyss&s|44? bb 0@0@0B0Bbbb!"c"c1A1A1C1C"c"c"cdd27799::fXE!444r<rct}|g}t|dkr|d}|jj|vrvt |t r=|jt|jjd|d|t|jz}|t| z }t|dkt|S)a Get the modules of the model that should not be spit when using device_map. We iterate through the modules to get the underlying `_no_split_modules`. Args: device_map (`str`): The device map value. Options are ["auto", "balanced", "balanced_low_0", "sequential"] Returns: `List[str]`: List of modules that should not be split rrNz does not support `device_map='z_'`. To implement support, the model class needs to implement the `_no_split_modules` attribute.) r1rXrrfrorr]_no_split_modulesrrUr)rerrkmodules_to_checkr.s r9_get_no_split_modulesz ModelMixin._get_no_split_moduless EE 6"##a''%))"--F(0AAAfj11^/7(%/8ZZYcZZZ ->FD\@]@],]) D):):$;$;; "##a''%&&&r<c t|S)z `torch.device`: The device on which the module is (assuming that all the module parameters are on the same device). )rSr{s r9rKzModelMixin.devices $D)))r<c t|S)zw `torch.dtype`: The dtype of the module (assuming that all the module parameters have the same dtype). )r[r{s r9rYzModelMixin.dtypes #4(((r<only_trainableexclude_embeddingsc|rYd|Dfd|D}tfd|DStfd|DS)a Get number of (trainable or non-embedding) parameters in the module. Args: only_trainable (`bool`, *optional*, defaults to `False`): Whether or not to return only the number of trainable parameters. exclude_embeddings (`bool`, *optional*, defaults to `False`): Whether or not to return only the number of non-embedding parameters. Returns: `int`: The number of parameters. Example: ```py from diffusers import UNet2DConditionModel model_id = "runwayml/stable-diffusion-v1-5" unet = UNet2DConditionModel.from_pretrained(model_id, subfolder="unet") unet.num_parameters(only_trainable=True) 859520964 ``` c\g|])\}}t|tjj$|d*S)z.weight)rr3r Embedding)r6rg module_types r9r:z-ModelMixin.num_parameters..sJ%%%%D+k58+=>>%   %%%r<c"g|] \}}|v | Sr1r1)r6rgr+embedding_param_namess r9r:z-ModelMixin.num_parameters..s/(((-dIDXmLmLm LmLmLmr<c3RK|]!}|js |V"dSrb requires_gradnumelr6prps r9rxz,ModelMixin.num_parameters..!s8llQaol]klqwwyyllllllr<c3RK|]!}|js |V"dSrbryr|s r9rxz,ModelMixin.num_parameters..#s8eeQqeVdeqwwyyeeeeeer<) named_modulesnamed_parameterssumrH)rerprqnon_embedding_parametersrws ` @r9num_parameterszModelMixin.num_parameterss2  f%%)-););)=)=%%% ! ((((151F1F1H1H((( $llll*Blllll leeee$//*;*;eeeee er<cgfdd|D]+}|d|vr||d||d<|d|vr||d||d<|d|vr||d||d<|d |vr||d ||d <|d |vr||d ||d <|d |vr||d ||d<|d|vr||d||d<|d|vr||d||d<-dS)Nct|dr|jr||D] \}}|dkr|n|d|}||!dS)N_from_deprecated_attn_blockrr)rmrrKnamed_children)rgr.sub_name sub_module deprecated_attention_block_pathsrecursive_find_attn_blocks r9rzRModelMixin._convert_deprecated_attention_blocks..recursive_find_attn_block(sv<== >&Bd >077===(.(=(=(?(? @ @$*'+rzz88$7K7K7K7K))(J???? @ @r<rz .query.weightz .to_q.weightz .query.biasz .to_q.biasz .key.weightz .to_k.weightz .key.biasz .to_k.biasz .value.weightz .to_v.weightz .value.biasz .to_v.biasz.proj_attn.weightz.to_out.0.weightz.proj_attn.biasz.to_out.0.bias)r)rerrrrs @@r9r0z/ModelMixin._convert_deprecated_attention_blocks%sK+-( @ @ @ @ @ @ "!"d+++ 5 _ _D%%%334>NNdCYCYCY4Z4Z d0001###z112<..DAUAUAU2V2V d.../###z114>NNdCWCWCW4X4X d0001!!!Z//2<..DASASAS2T2T d.../%%%334>NNdCYCYCY4Z4Z d0001###z112<..DAUAUAU2V2V d.../)))Z778B$GaGaGa8b8b d4445''':556@nnE]E]E]6^6^ d22233 _ _r<cgfd|D]@}|j|_|j|_|j|_|jd|_|`|`|`|`AdS)Nct|dr|jr||D] }|dSNrrmrrKrr.r"deprecated_attention_block_modulesrs r9rz_ModelMixin._temp_convert_self_to_deprecated_attention_blocks..recursive_find_attn_blockTmv<== B&Bd B299&AAA$oo// 6 6 ))*5555 6 6r<r)to_qqueryto_krRto_vrto_out proj_attnrer.rrs @@r9r5z.recursive_find_attn_blockorr<r,N) rrrRrrrr ModuleListrDropoutdropoutrrs @@r9r6zAModelMixin._undo_temp_convert_self_to_deprecated_attention_blocksls-/* 6 6 6 6 6 6 6 "!$'''8 ! !F ,FK *FK ,FKM6+;RZ=W=W*XYYFM      ! !r<rrb)TNTNrF)F)FF)/ro __module__ __qualname____doc__r config_name_automatically_saved_argsrr2rkrdrNrrppropertyboolr|rrrrrr rrrrr rPathLiker'r classmethodrrGrr7rirmr3rKrYrr0r5r6 __classcell__)rfs@r9r]r]tsK V V V',$)-&))))))))$n4nnnXn JJJJOOOO==$==== ////0000?C111)1();1 1111""M"MxPXGY"Mei"M"M"M"MH@@@@!%,0#'!%*0!NNc2;./NN ) N ! N # Nc3hNNNNN`CHU3PRP[K[E\<]CCC[CJ). eQeQ eQ (-S"+-='> eQ "& eQeQeQ[eQN55[5''''':* ***X*)u{)))X) $f$fT$ft$f`c$f$f$f$fL*_{*_t*_*_*_*_X6!!!!!!!!r<r]cdeZdZdZeedeeee j ffdZ dS)LegacyModelMixinz A subclass of `ModelMixin` to resolve class mapping from legacy classes (like `Transformer2DModel`) to more pipeline-specific classes (like `DiTTransformer2DModel`). rc ddlm}|}|dd}|dd}|dd}|dd}|dd} |d d} |d d} |d d} |} td d d}|j| f|dd|||| | | | |d |\}}}|||}|j|fi|S)Nr#)_fetch_remapped_cls_from_configrrFrrrrrrr r r Tr)model_loading_utilsrcopyrrr(rG)r;rrr kwargs_copyrrrrrrrrr=rr>_remapped_classs r9rGz LegacyModelMixin.from_pretrainedsc IHHHHHkkmm JJ{D11 $4e<< **%6==**Y--!::&8$?? 7D))::j$//JJ{D11 4 % "  's  !%#)+-!    1 98EE-~-.K[[{[[[r<N) rorrrrrr r rNrrrGr1r<r9rrsj ,\HU3PRP[K[E\<],\,\,\[,\,\,\r<r)Cr`rFrrr collectionsr functoolsrpathlibrtypingrrrr r r rr3huggingface_hubr r huggingface_hub.utilsrrrrrutilsrrrrrrrrrrrrrutils.hub_utilsr r!r"rr$r%r&r'r( get_loggerrorcompilerr#r.rMrKrSrYr[r]rr1r<r9rsF" ######>>>>>>>>>>>>>>>> KKKKKKKK666666   H % %rz122 D'""'!%!& %EHO % % % % % $58?$u{$$$$,M!M!M!M!M!.M!M!M!` 4\4\4\4\4\z4\4\4\4\4\r<