o s hE@sddlZddlZddlmZddlmZmZmZmZm Z m Z m Z ddl Z ddl mZmZmZmZde jdefddZde jdefd d Zde jdefd d Zd ee edeee jfdee efddZdeee jfdee efddZde jdefddZddddeee jfdeeedeeedeeeeffddZ d@de jjdedeeeefdefdd Z !dAde jjde eej fd"ed#e eefde eeeeff d$d%Z!dBd eee jfdeeeefde"fd&d'Z# dBd eee jfde eej fdeeeeffd(d)Z$dCde eej fd#e eefdeee jffd*d+Z%d,e"deee jffd-d.Z&e'e d/dZ(e'e d0dZ)e j*d1e j+d2e j,d2e j-d3e j.d3e j/d3e j0d4e j1d4e jd4e j2d1e(d4e)d4i Z3e j2e j+e j.e j-e j*e j,e j/e j1e j0e je(e)d5 Z4d6ede j5fd7d8Z6deee jffd9d:Z7de jd;ede"fdd?Z9dS)DN) defaultdict)AnyDictListOptionalSetTupleUnion) deserialize safe_open serializeserialize_filetensorreturnc CsHz|WSty#z |WYSty"YYdSww)Nr)untyped_storagedata_ptr ExceptionstorageNotImplementedErrorrre/var/www/html/construction_image-detection-poc/venv/lib/python3.10/site-packages/safetensors/torch.py storage_ptr s  rcCs4|r|ddt|j}|S|}|S)N)nelementviewr_SIZEdtype)rstoprrr_end_ptrs rc Csdz|WSty1z|t|jWYSty0|t|jYYSwwN) rnbytesAttributeErrorrsizerrrrrrrr storage_sizes  r$tensors state_dictc Csg}|D]V}t|dkr||qg}|D]}||}||t||fq||d\}}} || h|ddD]\} } }| |krP||hn|d|| }q@q|S)Nrr)lenappendrrsortadd) r%r&filtered_tensorssharedareasnamer_ last_stop last_namestartrrrr_filter_shared_not_shared,s&   r5cCs~tt}|D]'\}}|jtdkr/t|dkr/t|dkr/||jt|t|f|qtt | }t ||}|S)Nmetar) rsetitemsdevicetorchrr$r,listsortedvaluesr5)r&r%kvrrr_find_shared_tensorsEs( r@cCs*|t|ko|t|jt|kSr )rrrrrr$rrrr _is_completePs*rA)preferred_names discard_namesrBrCc s|durg}t|}|durg}t|}t}tt}|D]Q}tfdd|D}|s5td|dtt|d}||}|rLtt|d}|r]||}|r]tt|d}t|D] } | |krn||| qaq|S)Ncsg|] }t|r|qSr)rA).0r0r&rr dsz+_remove_duplicate_names..zvError while trying to find names to remove to save state dict, but found no suitable name to keep for saving amongst: z. None is covering the entire storage.Refusing to save/load the model since you could be storing much more memory than needed. Please refer to https://huggingface.co/docs/safetensors/torch_shared_tensors for more information. Or open an issue.r) r7r@rr; RuntimeErrorr< difference intersectionr*) r&rBrCshareds to_remover.complete_names keep_name preferredr0rrEr_remove_duplicate_namesTs<   rOTmodelfilenamemetadataforce_contiguousc Cs|}t|}|D]\}}|D]}|duri}||vr"|||<||=qq |r2dd|D}z t|||dWdStyT} z t| } | d7} t| d} ~ ww)a Saves a given torch model to specified filename. This method exists specifically to avoid tensor sharing issues which are not allowed in `safetensors`. [More information on tensor sharing](../torch_shared_tensors) Args: model (`torch.nn.Module`): The model to save on disk. filename (`str`): The filename location to save the file metadata (`Dict[str, str]`, *optional*): Extra information to save along with the file. Some metadata will be added for each dropped tensors. This information will not be enough to recover the entire shared structure but might help understanding things force_contiguous (`boolean`, *optional*, defaults to True): Forcing the state_dict to be saved as contiguous tensors. This has no effect on the correctness of the model, but it could potentially change performance if the layout of the tensor was chosen specifically for that reason. NcSsi|] \}}||qSr) contiguousrDr>r?rrr szsave_model..rRzT Or use save_model(..., force_contiguous=True), read the docs for potential caveats.)r&rOr8 save_file ValueErrorstr) rPrQrRrSr& to_removes kept_nameto_remove_grouprKemsgrrr save_models(r`cpustrictr9cCst||d}|}t||d}|j|dd\}}t|}|D]} | D]} | |vr3|| q'|| q'q#|rv|s@|rvd ddt |D} d ddt |D} d |j j d } |ri| d | 7} |rr| d | 7} t | ||fS) aA Loads a given filename onto a torch model. This method exists specifically to avoid tensor sharing issues which are not allowed in `safetensors`. [More information on tensor sharing](../torch_shared_tensors) Args: model (`torch.nn.Module`): The model to load onto. filename (`str`, or `os.PathLike`): The filename location to load the file from. strict (`bool`, *optional*, defaults to True): Whether to fail if you're missing keys or having unexpected ones. When false, the function simply returns missing and unexpected names. device (`Union[str, int]`, *optional*, defaults to `cpu`): The device where the tensors need to be located after load. available options are all regular torch device locations. Returns: `(missing, unexpected): (List[str], List[str])` `missing` are names in the model which were not modified during loading `unexpected` are names that are on the file, but weren't used during the load. )r9)rBF)rbz, cSg|]}d|dqS"rrDr>rrrrFzload_model..cSrcrdrrfrrrrFrgz#Error(s) in loading state_dict for :z# Missing key(s) in state_dict: z& Unexpected key(s) in state_dict: ) load_filer&rOkeysload_state_dictr7r=r*removejoinr< __class____name__rG)rPrQrbr9r&model_state_dictr[missing unexpectedr]rK missing_keysunexpected_keyserrorrrr load_models*     rvcCstt||d}t|}|S)a$ Saves a dictionary of tensors into raw bytes in safetensors format. Args: tensors (`Dict[str, torch.Tensor]`): The incoming tensors. Tensors need to be contiguous and dense. metadata (`Dict[str, str]`, *optional*, defaults to `None`): Optional text only metadata you might want to save in your header. For instance it can be useful to specify more about the underlying tensors. This is purely informative and does not affect tensor loading. Returns: `bytes`: The raw bytes representing the format Example: ```python from safetensors.torch import save import torch tensors = {"embedding": torch.zeros((512, 1024)), "attention": torch.zeros((256, 256))} byte_data = save(tensors) ``` rW)r _flattenbytes)r%rR serializedresultrrrsavesr{cCstt|||ddS)ah Saves a dictionary of tensors into raw bytes in safetensors format. Args: tensors (`Dict[str, torch.Tensor]`): The incoming tensors. Tensors need to be contiguous and dense. filename (`str`, or `os.PathLike`)): The filename we're saving into. metadata (`Dict[str, str]`, *optional*, defaults to `None`): Optional text only metadata you might want to save in your header. For instance it can be useful to specify more about the underlying tensors. This is purely informative and does not affect tensor loading. Returns: `None` Example: ```python from safetensors.torch import save_file import torch tensors = {"embedding": torch.zeros((512, 1024)), "attention": torch.zeros((256, 256))} save_file(tensors, "model.safetensors") ``` rWN)r rw)r%rQrRrrrrXsrXcCsTi}t|d|d}|D] }||||<qWd|S1s#wY|S)a Loads a safetensors file into torch format. Args: filename (`str`, or `os.PathLike`): The name of the file which contains the tensors device (`Union[str, int]`, *optional*, defaults to `cpu`): The device where the tensors need to be located after load. available options are all regular torch device locations. Returns: `Dict[str, torch.Tensor]`: dictionary that contains name as key, value as `torch.Tensor` Example: ```python from safetensors.torch import load_file file_path = "./my_folder/bert.safetensors" loaded = load_file(file_path) ``` pt) frameworkr9N)r rj get_tensor)rQr9rzfr>rrrri!s  ridatacCst|}t|S)a Loads a safetensors file into torch format from pure bytes. Args: data (`bytes`): The content of a safetensors file Returns: `Dict[str, torch.Tensor]`: dictionary that contains name as key, value as `torch.Tensor` on cpu Example: ```python from safetensors.torch import load file_path = "./my_folder/bert.safetensors" with open(file_path, "rb") as f: data = f.read() loaded = load(data) ``` )r _view2torch)rflatrrrload?sr float8_e4m3fn float8_e5m2r'r() F64F32F16BF16I64I32I16I8U8BOOLF8_E4M3F8_E5M2 dtype_strcCst|Sr )_TYPES)rrrr _getdtypesrcCsi}|D]K\}}t|d}t|ddkr-tdd|dDs#Jtj|d|d}ntj|d|d|d}tjdkrKt | j d d }|||<q|S) Nrrrcss|]}|dkVqdS)rNr)rDxrrr sz_view2torch..shape)rbigFinplace) rr)anyr:empty frombufferreshapesys byteorder from_numpynumpybyteswap)safeviewrzr>r?rarrrrrrs    rr0c CsD|jtjkrtd|d|std|d|jjdkr%|d}ddl}ddl }t | |j  }t|j}||}|}|dkrJdS||||j}|j||f} tjdkrtj|jtj|jtj|jtj|jtj|jtj|jtj|jtj|jtj t tj!|j!t"|jt#|ji } | |j} | $| j%d d } | &S) Nz)You are trying to save a sparse tensor: `` which this library does not support. You can make it a dense tensor before saving with `.to_dense()` but be aware this might make a much larger file than needed.z1You are trying to save a non contiguous tensor: `a` which is not allowed. It either means you are trying to save tensors which are reference of each other in which case it's recommended to save only the full tensors, and reslice at load time, or simply call `.contiguous()` on your tensor to pack it before saving.rarrFr)'layoutr:stridedrY is_contiguousr9typetoctypesrintprodritemrrrcastPOINTERc_ubyte ctypeslibas_arrayrrint64float32int32bfloat16float16int16uint8int8boolfloat64_float8_e4m3fn _float8_e5m2rrtobytes) rr0rnplengthbytes_per_item total_bytesptrnewptrrNPDTYPESnpdtyperrr_tobytessJ        rcCst|tstdt|g}|D]!\}}t|tjs*td|dt||jtjkr5| |q|r@td|dt |}g}|D] }t |dkrU| |qH|r`t d|dd d |DS) Nz4Expected a dict of [str, torch.Tensor] but received zKey `z1` is invalid, expected torch.Tensor but received z*You are trying to save a sparse tensors: `rr(z Some tensors share memory, this will lead to duplicate memory on disk and potential differences when loading them again: z. A potential way to correctly save your model is to use `save_model`. More information at https://huggingface.co/docs/safetensors/torch_shared_tensors cSs6i|]\}}|t|jdd|jt||dqS).r)rrr)rZrsplitrrrUrrrrVsz_flatten..) isinstancedictrYrr8r:Tensorrrr*r@r)rG)r%invalid_tensorsr>r?shared_pointersfailingnamesrrrrws8       rw)NT)Trar )ra):osr collectionsrtypingrrrrrrr r: safetensorsr r r r rrrrr$rZr5r@rrArOnnModuler`PathLikervrxr{rXrirgetattrrrrrrrrrrrrrrrrrrrwrrrrs $ 0 $     / /   .1!   4"  0: