o پi@sddlZddlZddlZddlZddlmZmZddlZddlm Z m Z ddl m Z m Z mZmZdaGddde Zedd d Zdd eefddZddZddZejddZdS)N)CallableOptional)CacheDisk)ENOVALUNKNOWN args_to_key full_nameTcsReZdZd fdd ZfddZfddZeffdd Zfd d ZZ S) CloudpickleDiskc s||_tj|fi|dSN)compress_levelsuper__init__)self directoryr kwargs __class__D/home/ubuntu/.local/lib/python3.10/site-packages/outlines/caching.pyrszCloudpickleDisk.__init__cst|}t|Sr ) cloudpickledumpsrput)rkeydatarrrrs  zCloudpickleDisk.putcst||}t|Sr )rgetrloads)rrrawrrrrrs zCloudpickleDisk.getcs |st|}tj|||dS)N)r)rrrstore)rvaluereadrrrrrs zCloudpickleDisk.storecs$t||||}|st|}|Sr )rfetchrr)rmodefilenamer r!rrrrr" s zCloudpickleDisk.fetch)r ) __name__ __module__ __qualname__rrrrrr" __classcell__rrrrr s   r r cCs^ddlm}tjd}tjd|d}t|ddtd}||dkr)| ||d<|S) aGet the context object that contains previously-computed return values. The cache is used to avoid unnecessary computations and API calls, which can be long and expensive for large models. The cache directory defaults to `HOMEDIR/.cache/outlines`, but this choice can be overridden by the user by setting the value of the `OUTLINES_CACHE_DIR` environment variable. r) __version__~OUTLINES_CACHE_DIRz/.cache/outlinesnone)eviction_policy cull_limitdiskr)) outlines._versionr)ospath expanduserenvironrrr clear)outlines_versionhome_dir cache_dirmemoryrrr get_cache's r:Frexpirecsdtffdd }|S)a4Caching decorator for memoizing function calls. The cache key is created based on the values returned by the key_function callable if provided or based on the arguments of the decorated function directly otherwise This is based on `diskcache`'s `memoize`. Parameters ---------- expire Seconds until arguments expire. typed Cache different types separately. ignore Positional or keyword arguments to ignore. Returns ------- A decorator function that can be applied to other functions. cached_functioncsbt}tftrfddnfddfdd}|_|__S)Ncspts |i|IdHSj|i|}jj|tdd}|tur6|i|IdH}jj||dd|SNT)defaultretry)r?_caching_enabled __cache_key__ __memory__rrsetargsr cache_keyresultr<r;wrapperrrrJcsz)cache..decorator..wrappercsbts |i|Sj|i|}jj|tdd}|tur/|i|}jj||dd|Sr=r@rErIrrrJrscst||S)z,Make key for cache given function arguments.)r)rFr)baseignoretypedrrrBsz/cache..decorator..__cache_key__)r:r asyncioiscoroutinefunctionrBrC __wrapped__)r<r9rBr;rLrM)rKr<rJr decorator\s   zcache..decorator)r)r;rMrLrRrrQrcacheFs-rScCsdadS)auDisable the cache for this session. Generative models output different results each time they are called when sampling. This can be a desirable property for some workflows, in which case one can call `outlines.call.disable` to disable the cache for the session. This function does not delete the cache, call `outlines.cache.clear` instead. It also does not overwrite the cache with the values returned during the session. Example ------- `outlines.cache.disable` should be called right after importing outlines: >>> import outlines.caching as cache >>> cache.disable_cache() FNrArrrr disable_cachesrUcCst}|dS)zErase the cache completely.N)r:r5)r9rrr clear_caches rVccs"t}dazdVW|adS|aw)NFrT)original_staterrrcache_disableds rX)NFr)rN contextlib functoolsr1typingrrr diskcacherrdiskcache.corerrrr rAr lru_cacher:floatrSrUrVcontextmanagerrXrrrrs" F