Xc@`s/ddlmZmZmZddlZddlZddlZddlZddladdl Z ddl Z ddl Z ddl Z ddl Z ddlZddlmZddlZddlmZmZmZmZmZmZmZmZmZddlmZddlmZm Z ddl!m"Z"ej#dZ$eZ%d Z&ed d ed d dde&de'eddeddddde(eddeddeddeddd de(defdYZ)ed d!e)d"d#e(de(ed$d%e)d&d#e(de(ed'd(ee(d#e(de(ed)d*ee(de(ed+d,ee(de(d-efd.YZ*ed/d0e*de(ed1d2ee'd#e(de(ed3d4ee'd#e(de(d5Z+ed6d7ee+de(ed8d9ee'de(d:Z,ed;d<ed&e,de(ed=d>ed&de(ed?d&ee(de(ed@dAedBdCde(edDdEee(de'edFdGedd#e(de(edHdIedJdKdLedMdNee'de(dOZ-edPdQed&d#e(dRe-de(dSZ.edTdUed&d#e(dRe.de(dVZ/dZ0dZ1dZ2edbdced&d#e(dRe/de(edddeee0de(edfdgee1de(edhdiee2de(edjdkedldmdd d de(dnZ3edodpee3dqde(edrdpee3ej dskrdtndude(edvdwedxdydzde(ed{d|eddddd#e'de(ed}d~eddddddde(dZ4yeddgZ5Wne6k rhdZ5nXe5dkr+ej dkr+ej7j8ej7j9ej:ddddZ;y>ee;dgZ5e5dkre"dej7j9e;nWne6k rdZ5nXe5dkr(e$j<dq(q+ne5dkr@d&Z4nej dskry+eddgZ5e5dkrydZ4nWqe6k rqXne4d&krddl=Z>e>j?j@e4ZAeAdk reAZ4n[A[>ne4rejCdkrde4Z4ndZDeddee4deDde([4e5dkrxe%jEd&krxeddeddddddddde(nDeddedddde(eFe%jEjGre$j<dneddee'de(eddeddddde(eddee(de(eddeddddde(dZHedded&d#e(dReHde(eddee(de(eddedddde(eddeddd#e(de(eddee'de(eddee'de(eddee(de(eddee'de(eddedde(eddedde(eddedJe(de(eddedddddddd#e(de(eddedddddddd#e(de(eddedddddddd#e(de(eddedddddddd#e(de(eddedddddddd#e(de(eddedddddddddddddd#e( de(dZIeddeeIdde(eddeeIdde(eddeeIdde(eddeeIdde(eddeeIdde(eddeeIdde(eddee(de(eddeeIdde(eddeeIdde(eddeeIdde(eddeeIdde(eddedddddde(eddee(de(eddedddddde(eddee'de(eddee(d#e'de(eddeddde(ejJddZKeKryeLeKWn!eMk rqeNdeKqXeLeKdk ZOn1eZPePdkre$j<d nePdkZOe(ZOed d eeOde(ed d edde(eddee'de'eddee(de(dZQeddeddeQde(eddee'de(eddee'de(eddee'de(eddedddde(edd ed&d#e(de(ed!d"ed&d#e(de(ed#d$ed&d#e(de(ed%d&ed'd(de(ed)d*ed+de(ed,d-ee'de(ed.d/ee'de(ed0d1edd2de(ed3d4ee'de(d5ZRed6d7ed&deRde(ed8d9ed:d;de(ed<d=ee'de(ed>d?ed@dAde(edBdCed@dDde(edEdFedGdHde(edIdJedKdLde(edMdNee(de(edOdPedQde(edRdSee(de(edTdUee(de(edVdWetjSde(edXdYedde(edZd[ed&de(ed\d]ee(de(ed^d_ee(de'ed`daee(de(edbdcee(d#e(de(edddeedd#e(de(dhZTdiZUedjdkeeTde(edldmejVje(de(edndoee(de(edpdqee(de(edrdsee(de(dtZWedudvedeWde(edwdxeeIdde(edydzejVjddde(ed{d|ejVjdde(ed}d~ejVje'ddde(eddejVje'de(eddee(de(eddee'de(eddee(de(eddee(de(eddedde(eddeddd#e(de(eddeddd#e(de(dZXeddeeXdd#e(de(y/e e%jEdgZYeYdjZj[Z\Wne6k rdZ\nXdZ]dZ^i e j d6e j_d6e j`d6e]d6e^d6ejad6tjad6e\jbddd6e jcd6ZddddZeeeeddt;s->isMalformed context map: %siR<R(R'sCannot use %s as context name(tsplittlenR+R*(R-tvR((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyR/s #R>(RR4tFalse(R.R/((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyR4s (R9R:R4(((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRAstcontextssa Context map for multi-gpu operation. Format is a semicolon-separated list of names and device names in the 'name->dev_name' format. An example that would map name 'test' to device 'cuda0' and name 'test2' to device 'opencl0:0' follows: "test->cuda0;test2->opencl0:0". Invalid context names are 'cpu', 'cuda*' and 'opencl*' tprint_active_devices:Print active device at when the GPU device is initialized.tenable_initial_driver_tests9Tests the nvidia driver when a GPU device is initialized.cC`stjdd}|r|Stjd}|s5dSxZ|jtjjD]C}tjjtjj|drKtjjtjj|SqKWdS(Nt CUDA_ROOTR>tPATHtnvcc( tostgetenvRCtpathtpathseptexiststjointdirnametabspath(RERtdir((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytdefault_cuda_roots! s cuda.rootsdirectory with bin/, lib/, include/ for cuda utilities. This directory is included via -L and -rpath when linking dynamically compiled modules. If AUTO and nvcc is in the path, it will use one of nvcc parent directory. Otherwise /usr/local/cuda will be used. Leave empty to prevent extra linker directives. Default: environment variable "CUDA_ROOT" or else "AUTO". s cuda.enableds0If false, C code in old backend is not compiled.cC`st|tstg|jdD]}|r%|^q%}tg|D]}|jdsG|^qGr~td|ndj|S(Nt t-sTheano nvcc.flags support only parameter/value pairs without space between them. e.g.: '--machine 64' is not supported, but '--machine=64' is supported. Please add the '=' symbol. nvcc.flags value is '%s'(t isinstancetstrtAssertionErrorRCtanyR*R+RR(Rtflagtflagstf((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytfilter_nvcc_flagss(. s nvcc.flagssExtra compiler flags for nvccsnvcc.compiler_bindirsHIf defined, nvcc compiler driver will seek g++ and gcc in this directorys nvcc.fastmaths nvcc.cudafes If 'always' (the default), cudafe will be called for every GPU Op compilation. If 'heuristic', it will only be called if the source code appears to contain CUDA code. This can speed up compilation and importing theano, but might fail to compile some custom GPU Ops.talwayst heuristics gpuarray.syncsIf True, every op will make sure its work is done before returning. Setting this to True will slow down execution, but give much more accurate results in profiling.sgpuarray.preallocatesIf negative it disables the allocation cache. If between 0 and 1 it enables the allocation cache and preallocates that fraction of the total GPU memory. If 1 or greater it will preallocate that amount of memory (in megabytes).sgpuarray.schedstThe sched parameter passed for context creation to pygpu. With CUDA, using "multi" is equivalent to using the parameter cudaDeviceScheduleYield. This is useful to lower the CPU overhead when waiting for GPU. One user found that it speeds up his other processes that was doing data augmentation. R)tmultitsinglesgpuarray.single_streams If your computations are mostly lots of small elements, using single-stream will avoid the synchronization overhead and usually be faster. For larger elements it does not make a difference yet. In the future when true multi-stream is enabled in libgpuarray, this may change. If you want to make sure to have optimal performance, check both options. cC`s|rtdntS(sH Make sure the user is not attempting to use dnn.conv.workmem`. sThe option `dnn.conv.workmem` has been removed and should not be used anymore. Please use the option `dnn.conv.algo_fwd` instead.(t RuntimeErrorR2(tworkmem((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytsafe_no_dnn_workmem+s sdnn.conv.workmems/This flag is deprecated; use dnn.conv.algo_fwd.R/cC`s|rtdntS(sL Make sure the user is not attempting to use dnn.conv.workmem_bwd`. sThe option `dnn.conv.workmem_bwd` has been removed and should not be used anymore. Please use the options `dnn.conv.algo_bwd_filter` and `dnn.conv.algo_bwd_data` instead.(ReR2(Rf((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytsafe_no_dnn_workmem_bwd<s sdnn.conv.workmem_bwds]This flag is deprecated; use `dnn.conv.algo_bwd_filter` and `dnn.conv.algo_bwd_data` instead.cC`s|rtdntS(sI Make sure the user is not attempting to use dnn.conv.algo_bwd`. sThe option `dnn.conv.algo_bwd` has been removed and should not be used anymore. Please use the options `dnn.conv.algo_bwd_filter` and `dnn.conv.algo_bwd_data` instead.(ReR2(talgo((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytsafe_no_dnn_algo_bwdOs tsmalltnonetlargetfftt fft_tilingtwinogradt guess_oncetguess_on_shape_changet time_oncettime_on_shape_changet deterministicsdnn.conv.algo_bwdsQThis flag is deprecated; use dnn.conv.algo_bwd_data and dnn.conv.algo_bwd_filter.sdnn.conv.algo_fwds<Default implementation to use for cuDNN forward convolution.sdnn.conv.algo_bwd_datasDefault implementation to use for cuDNN backward convolution to get the gradients of the convolution with regard to the inputs.sdnn.conv.algo_bwd_filtersDefault implementation to use for cuDNN backward convolution to get the gradients of the convolution with regard to the filters.sdnn.conv.precisionsDefault data precision to use for the computation in cuDNN convolutions (defaults to the same dtype as the inputs of the convolutions, or float32 if inputs are float16).t as_input_f32tas_inputcC`s|d}|S(NcS`s5tjjjdkrdStjjtjjj|S(NR>(ttheanotconfigR(trootRMRORR(tsuffix((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyR_s((R{R_((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytdefault_dnn_paths sdnn.include_paths8Location of the cudnn header (defaults to the cuda root)tincludesdnn.library_pathtdarwintlibtlib64s dnn.enableds'auto', use cuDNN if available, but silently fall back to not using it if not present. If True and cuDNN can not be used, raise an error. If False, disable cudnntautoR2RFtassert_no_cpu_opsGRaise an error/warning if there is a CPU op in the computational graph.tmodesDefault compilation modetModet DebugModetFAST_RUNt NanGuardModet FAST_COMPILEt DEBUG_MODEsg++s-vitwin32tLibrarys mingw-w64tbinRKsBg++ not available, if using conda: `conda install m2w64-toolchain`sclang++tnts"%s"cC`s/tjdkr+d|kr+tjdntS(sAWe only support clang++ as otherwise we hit strange g++/OSX bugs.R~sclang++sIOnly clang++ is supported. With g++, we end up with strange g++/OSX bugs.(tsystplatformt_loggertwarningR2(R-((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytwarn_cxxstcxxsThe C++ compiler to use. Currently only g++ is supported, but supporting additional compilers should not be too difficult. If it is empty, no C++ code is compiled.tis_validtlinkers4Default linker used if the theano flags mode is Modetcvmsc|pytpytcs c|py_nogctvmtvm_nogctcvm_nogcsg++ not detected ! Theano will be unable to execute optimized C-implementations (for both CPU and GPU) and will default to Python implementations. Performance will be severely degraded. To remove this warning, set Theano flags cxx to an empty string.tallow_gcsDo we default to delete intermediate results during Theano function calls? Doing so lowers the memory requirement, but asks that we reallocate memory at the next function call. This is implemented for the default linker, but may not work for all linkers.t optimizersEDefault optimizer. If not None, will use this optimizer with the Modetfast_runtmerget fast_compiletNonetoptimizer_verboses0If True, we print all optimization being appliedt on_opt_errorsnWhat to do when an optimization crashes: warn and skip it, raise the exception, or fall into the pdb debugger.cC`s,|r(tdtjj|dntS(s Make sure the user is not attempting to use `config.home`. This config option was removed in Thenao 0.5 since it was redundant with `config.base_compiledir`. This filter function ensures people who were setting the location of their compilation directory through `config.home` switch to `config.basecompiledir` instead, by raising an error when `config.home` is used. sThe `config.home` option has been removed and should not be used anymore. Please set the `config.base_compiledir` option instead (for instance to: %s)s.theano(ReRMRORRR2(thome((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyt safe_no_home1s Rs5This config option was removed in 0.5: do not use it!t nocleanups@Suppress the deletion of code files that did not compile cleanlyton_unused_inputs]What to do if a variable in the 'inputs' list of theano.function() is not used in the graph.stensor.cmp_sloppys:Relax tensor._allclose (0) not at all, (1) a bit, (2) morecC`s |dkS(Niii(iii((ti((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyt^sstensor.local_elemwise_fusionsVEnable or not in fast_run mode(fast_run optimization) the elemwise fusion optimizationsgpu.local_elemwise_fusionsZEnable or not in fast_run mode(fast_run optimization) the gpu elemwise fusion optimizations lib.amdlibms#Use amd's amdlibm numerical librarysgpuelemwise.syncsBwhen true, wait that the gpu fct finished and check it error code.straceback.limits*The number of stack to trace. -1 mean all.istraceback.compile_limitsThe number of stack to trace to keep during compilation. -1 mean all. If greater then 0, will also make us save Theano internal stack trace.s experimental.unpickle_gpu_on_cpusAllow unpickling of pickled CudaNdarrays as numpy.ndarrays.This is useful, if you want to open a CudaNdarray without having cuda installed.If you have cuda installed, this will force unpickling tobe done on the cpu to numpy.ndarray.Please be aware that this may get you access to the data,however, trying to unpicke gpu functions will not succeed.This flag is experimental and may be removed any time, whengpu<>cpu transparency is solved.snumpy.seterr_alls2Sets numpy's behaviour for floating-point errors, sDsee numpy.seterr. 'None' means not to change numpy's default, which can be different for different numpy releases. This flag sets the default behaviour for all kinds of floating-point errors, its effect can be overriden for specific errors by the following flags: seterr_divide, seterr_over, seterr_under and seterr_invalid.tcalltprinttlogsnumpy.seterr_divideszSets numpy's behavior for division by zero, see numpy.seterr. 'None' means using the default, defined by numpy.seterr_all.snumpy.seterr_oversSets numpy's behavior for floating-point overflow, see numpy.seterr. 'None' means using the default, defined by numpy.seterr_all.snumpy.seterr_undersSets numpy's behavior for floating-point underflow, see numpy.seterr. 'None' means using the default, defined by numpy.seterr_all.snumpy.seterr_invalidsSets numpy's behavior for invalid floating-point operation, see numpy.seterr. 'None' means using the default, defined by numpy.seterr_all.swarn.ignore_bug_beforesIf 'None', we warn about all Theano bugs found by default. If 'all', we don't warn about Theano bugs found by default. If a version, we print only the warnings relative to Theano bugs found after that version. Warning for specific bugs can be configured with specific [warn] flags.s0.7talls0.3s0.4s0.4.1s0.5s0.6s0.8s0.8.1s0.8.2s0.9cC`sFtjjdkrtStjjdkr,tStjj|krBtStS(sP Return True iff we should warn about bugs fixed after a given version. RR(RyRtignore_bug_beforeR2RF(tversion((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyt warn_defaultsswarn.argmax_pushdown_bugsWarn if in past version of Theano we generated a bug with the theano.tensor.nnet.nnet.local_argmax_pushdown optimization. Was fixed 27 may 2010swarn.gpusum_01_011_0111_bugsWarn if we are in a case where old version of Theano had a silent bug with GpuSum pattern 01,011 and 0111 when the first dimensions was bigger then 4096. Was fixed 31 may 2010swarn.sum_sum_bugsWarn if we are in a case where Theano version between version 9923a40c7b7a and the 2 august 2010 (fixed date), generated an error in that case. This happens when there are 2 consecutive sums in the graph, bad code was generated. Was fixed 2 August 2010swarn.sum_div_dimshuffle_bugsWarn if previous versions of Theano (between rev. 3bd9b789f5e8, 2010-06-16, and cfc6322e5ad4, 2010-08-03) would have given incorrect result. This bug was triggered by sum of division of dimshuffled tensors.swarn.subtensor_merge_bugsWarn if previous versions of Theano (before 0.5rc2) could have given incorrect results when indexing into a subtensor with negative stride (for instance, for instance, x[a:b:-1][c]).swarn.gpu_set_subtensor1sWarn if previous versions of Theano (before 0.6) could have given incorrect results when moving to the gpu set_subtensor(x[int vector], new_value)swarn.vm_gc_bugsThere was a bug that existed in the default Theano configuration, only in the development version between July 5th 2012 and July 30th 2012. This was not in a released version. If your code was affected by this bug, a warning will be printed during the code execution if you use the `linker=vm,vm.lazy=True,warn.vm_gc_bug=True` Theano flags. This warning is disabled by default as the bug was not released.swarn.signal_conv2d_interfacesLWarn we use the new signal.conv2d() when its interface changed mid June 2014swarn.reduce_joinsYour current code is fine, but Theano versions prior to 0.7 (or this development version) might have given an incorrect result. To disable this warning, set the Theano flag warn.reduce_join to False. The problem was an optimization, that modified the pattern "Reduce{scalar.op}(Join(axis=0, a, b), axis=0)", did not check the reduction axis. So if the reduction axis was not 0, you got a wrong answer.swarn.inc_set_subtensor1sWarn if previous versions of Theano (before 0.7) could have given incorrect results for inc_subtensor and set_subtensor when using some patterns of advanced indexing (indexing with one vector or matrix of ints).s warn.roundsxRound changed its default from Seed to use for randomized unit tests. Special value 'random' means using a seed of None.tcompute_test_valuesIf 'True', Theano will run each op at graph build time, using Constants, SharedVariables and the tag 'test_value' as inputs to the function. This helps the user track down problems in the graph before it gets optimized.tofftprint_test_valuesIf 'True', the __eval__ of a Theano variable will return its test_value when this is available. This has the practical conseguence that, e.g., in debugging `my_var` will print the same as `my_var.tag.test_value` when a test value is defined.tcompute_test_value_optsjFor debugging Theano optimization only. Same as compute_test_value, but is used during Theano optimizationtunpickle_functionswReplace unpickled Theano functions with None. This is useful to unpickle old graphs that pickled them when it shouldn'ttreoptimize_unpickled_functionsHRe-optimize the graph when a theano function is unpickled from the disk.texception_verbositys7If 'low', the text of exceptions will generally refer to apply nodes with short names such as Elemwise{add_no_inplace}. If 'high', some exceptions will also refer to apply nodes with long descriptions like: A. Elemwise{add_no_inplace} B. log_likelihood_v_given_h C. log_likelihood_htlowthightOMP_NUM_THREADSsFThe environment variable OMP_NUM_THREADS should be a number, got '%s'.isWe are not able to detect the number of CPU cores. We disable openmp by default. To remove this warning, set the environment variable OMP_NUM_THREADS to the number of threads you want theano to use.topenmpsPAllow (or not) parallel computation on the CPU with OpenMP. This is the default value used when creating an Op that supports OpenMP parallelization. It is preferable to define it via the Theano configuration file ~/.theanorc or with the environment variable THEANO_FLAGS. Parallelization is only done for some operations that implement it, and even for operations that implement parallelism, each operation is free to respect this flag or not. You can control the number of threads used with the environment variable OMP_NUM_THREADS. If it is set to 1, we disable openmp in Theano by default.topenmp_elemwise_minsizes~If OpenMP is enabled, this is the minimum size of vectors for which the openmp parallelization is enabled in element wise ops.i@ t check_inputsSpecify if types should check their input in their C code. It can be used to speed up compilation, reduce overhead (particularly for scalars) and reduce the number of generated C files.tcache_optimizationssWARNING: work in progress, does not work yet. Specify if the optimization cache should be used. This cache will any optimized graph and its optimization. Actually slow downs a lot the first optimization, and could possibly still contains some bugs. Use at your own risks.cC`s7|dkrtSyt|Wntk r2tSXtS(Ntrandom(R2R$t ExceptionRF(tseed((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytgood_seed_params  sunittests.rseedsYSeed to use for randomized unit tests. Special value 'random' means using a seed of None.isNanGuardMode.nan_is_errorsDefault value for nan_is_errorsNanGuardMode.inf_is_errorsDefault value for inf_is_errorsNanGuardMode.big_is_errorsDefault value for big_is_errorsNanGuardMode.actions.What NanGuardMode does when it finds a problemtoptimizer_excludings^When using the default mode, we will remove optimizer with these tags. Separate tags with ':'.toptimizer_includings[When using the default mode, we will add optimizer with these tags. Separate tags with ':'.toptimizer_requirings_When using the default mode, we will require optimizer with these tags. Separate tags with ':'.sDebugMode.patiences6Optimize graph this many times to detect inconsistencyi cC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRssDebugMode.check_cs$Run C implementations where possiblecC`sttjjS(N(tboolRxRyR(((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRssDebugMode.check_pys)Run Python implementations where possiblesDebugMode.check_finites&True -> complain about NaN/Inf resultssDebugMode.check_stridesswCheck that Python- and C-produced ndarrays have same strides. On difference: (0) - ignore, (1) warn, or (2) raise errorcC`s |dkS(Niii(iii((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRssDebugMode.warn_input_not_reusedsGenerate a warning when destroy_map or view_map says that an op works inplace, but the op did not reuse the input for its output.cC`s_t|tstSddddddddg}x'|jd D]}||krAtSqAWtS( Ntinitialtprevioust c_contiguoust f_contiguouststridedt wrong_sizetALLR>t:(RYRRFRCR2(tparamtvalidtp((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyt(is_valid_check_preallocated_output_params  s#DebugMode.check_preallocated_outputstTest thunks with pre-allocated memory as output storage. This is a list of strings separated by ":". Valid values are: "initial" (initial storage in storage map, happens with Scan),"previous" (previously-returned memory), "c_contiguous", "f_contiguous", "strided" (positive and negative strides), "wrong_size" (larger and smaller dimensions), and "ALL" (all of the above).s(DebugMode.check_preallocated_output_ndimsWhen testing with "strided" preallocated output memory, test all combinations of strides over that number of (inner-most) dimensions. You may want to reduce that number to reduce memory or time usage, but it is advised to keep a minimum of 2.icC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyR9ssprofiling.time_thunkss%Time individual thunks when profilingsprofiling.n_applys-Number of Apply instances to print by defaulticC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRCssprofiling.n_opss!Number of Ops to print by defaultcC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRHssprofiling.output_line_widths'Max line width for the profiling outputicC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRMssprofiling.min_memory_sizesFor the memory profile, do not print Apply nodes if the size of their outputs (in bytes) is lower than this thresholdicC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRSssprofiling.min_peak_memorys&The min peak memory usage of the ordersprofiling.destinationsD File destination of the profiling output tstderrsprofiling.debugprintsE Do a debugprint of the profiled functions sprofiling.ignore_first_callsM Do we ignore the first call of a Theano function. soptdb.position_cutoffsdWhere to stop eariler during optimization. It represent the position of the optimizer where to stop.soptdb.max_use_ratios;A ratio that prevent infinite loop in EquilibriumOptimizer.s gcc.cxxflagssExtra compiler flags for gccscmodule.warn_no_versionsIf True, will print a warning when compiling one or more Op with C code that can't be cached because there is no c_code_cache_version() function associated to at least one of those Ops.scmodule.remove_gxx_optsIf True, will remove the -O* parameter passed to g++.This is useful to debug in gdb modules compiled by Theano.The parameter -g is passed by default to g++scmodule.compilation_warnings)If True, will print compilation warnings.scmodule.preload_caches>If set to True, will preload the C module cache at import timescmodule.age_thresh_usesGIn seconds. The time after which Theano won't reuse a compile c module.i<ic C`sg}yttjdr9tjjr9tjjj}nLddlatjdt,dtjj j _ tjj j d}WdQXdt j krt}t jdkrdjdtjjt jd ggd d d gD]}d |^qSt jdkrtjjt jd}tjdd}|dk rMtjj|}n|dksn||jdkrtjd|t}qn|rdjdtjjt jdgg|dD]}d |^qSndt jkrd}t jdkrd }ntjjt j|}tjj|sRtjjt jddd}tjj|sltdtj|} xB| D]:} tjjtjj|| |s| j| qqWt| dkstd|| g| D]$} tjtjj|| ^qftjj|| d|}tjj|sRtdnt jdkspt jdkrdjd|gg|dD]}d |^qSt jdkrdjd|ggd d d gD]}d |^qSnyddl } WnWt!k rV} t"gd0D]} | t j kr| ^qr]|j#d| fq]nXt jdkrtjjt jdd g}d|g}n2|j$d!g}g}|rd|dg}n|gd"d#d$gD]}d |^q7}t%|}|r|S|j&g|j$d!gD]}d%|^qt%|}|r]t'd&|d|Stj(d'krrd(nd)}g|j$d!gD]}d*|||f^qg|j$dgD]}d |^q|j$d+g}t"d,|Dr |j&d-d-gnt%|}|r#|Sx|D]}tj|q*W~|j&g|j$d!gD]}d%|^q]t%|}|r|St jd1krtjjt jd}|j#d%|t%|}|r|SnWnt)k rnXt%d/gS(2Nt __config__itrecordtblas_opttEPDRRWs-L"%s"tScriptstmk2_coretmk2_intel_threadtmk2_rts-l%sR~RtDYLD_FALLBACK_LIBRARY_PATHRsThe environment variable 'DYLD_FALLBACK_LIBRARY_PATH' does not contain the '%s' path in its value. This will make Theano use a slow version of BLAS. Update 'DYLD_FALLBACK_LIBRARY_PATH' to contain the said value, this will disable this warning.s-L%st librariestCanopys..tappdatas"Canopy changed the location of MKLis5Unexpected case when looking for Canopy MKL librariestlinux2tcondat Continuums0install mkl with `conda install mkl-service`: %sRRt library_dirstmkl_coretmkl_intel_threadtmkl_rts -Wl,-rpath,RKRs"R>s-L%s%s%stextra_link_argscs`s|]}d|kVqdS(tmklN((t.0tfl((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pys ?ss-lmtlinuxs-lblas(RR(slinuxsdarwin(*thasattrtnumpyt distutilsRt blas_opt_infotnumpy.distutils.system_infotwarningstcatch_warningsR2t system_infot verbositytget_infoRRRRRRMROtprefixRNRtrealpathRCRRRFt base_prefixRQR[tlistdirtremoveRDRt ImportErrorR\tappendR1t try_blas_flagtextendRtnametKeyError(t warn_recordt blas_infot use_unix_epdtltnew_pathREtsubsubtlib_pathRt lib_pathstsubRtetmR^trest path_wrappertretR((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytdefault_blas_ldflagss    ! &  $ 4! #  ".   ' V   '    cC`sddlm}tjd}t|}tjdkr@dnd}|jgtj j j D]}d|||f^q_|j |dd d |d t }|r|dr|d rd j|SdSdS(Ni(t GCC_compilers extern "C" double ddot_(int*, double*, int*, double*, int*); int main(int argc, char** argv) { int Nx = 5; int Sx = 1; double x[5] = {0, 1, 2, 3, 4}; double r = ddot_(&Nx, x, &Sx, x, &Sx); if ((r - 30.) > 1e-6 || (r - 30.) < -1e-6) { return -1; } return 0; } Rs"R>s-L%s%s%st tmp_prefixt try_blas_R^ttry_runiRW(ttheano.gof.cmoduleRttextwraptdedenttlistRMRRRxtgoftcmodulet std_lib_dirsttry_compile_tmpR2RR(R^Rt test_codetcflagsR tdR ((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRis  9  s blas.ldflagss;lib[s] to include for [Fortran] level-3 blas implementationsmetaopt.verboses)Enable verbose output for meta optimizerstprofiles(If VM should collect profile informationtprofile_optimizers2If VM should collect optimizer profile informationtprofile_memorys<If VM should collect memory profile information and print itcC`sh|dks|tkrtS|dks4|tkr8tS|dksP|dkrTdStd|dS(NRFR2RsNValid values for an vm.lazy parameter should be None, False or True, not `%s`.(RFR2RR+(R-((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytfilter_vm_lazyssvm.lazysUseful only for the vm linkers. When lazy is None, auto detect if lazy evaluation is needed and use the apropriate version. If lazy is True/False, force the version used between Loop/LoopGC and Stack.swarn.identify_1pexp_bugs}Warn if Theano versions prior to 7987b51 (2011-12-18) could have yielded a wrong result due to a bug in the is_1pexp functionton_shape_errorsFwarn: print a warning and use the default value. raise: raise an errors+tensor.insert_inplace_optimizer_validate_nbs6-1: auto, if graph have less then 500 nodes 1, else 10s!experimental.local_alloc_elemwisesDEPRECATED: If True, enable the experimental optimization local_alloc_elemwise. Generates error if not True. Use optimizer_excluding=local_alloc_elemwise to dsiable.cC`s|S(N((tx((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRss(experimental.local_alloc_elemwise_assertsRWhen the local_alloc_elemwise is applied, add an assert to highlight shape errors.s scan.allow_gcs1Allow/disallow gc inside of Scan (default: False)sscan.allow_output_preallocsNAllow/disallow memory preallocation for outputs inside of scan (default: True)s scan.debugs4If True, enable extra verbose output related to scans pycuda.initsIf True, always initialize PyCUDA when Theano want to initilize the GPU. Currently, we must always initialize PyCUDA before Theano do it. Setting this flag to True, ensure that, but always import PyCUDA. It can be done manually by importing theano.misc.pycuda_init before theano initialize the GPU device. s cublas.libs-Name of the cuda blas library for the linker.tcublass lib.cnmemsiDo we enable CNMeM or not (a faster CUDA memory allocator). The parameter represent the start size (in MB or % of total GPU memory) of the memory pool. 0: not enabled. 0 < N <= 1: % of the total GPU memory (clipped to .985 for driver memory) > 0: use that number of MB of memory. cC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyR ss compile.waits8Time to wait before retrying to aquire the compile lock.icC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRscC`stjjjdS(Ni(RxRytcompiletwait(((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyt_timeout_defaultsscompile.timeouts-In seconds, time that a process will wait before deciding to override an existing lock. An override only happens when the existing lock is held by the same owner *and* has not been 'refreshed' by this owner for more than this period. Refreshes are done every half timeout period for running processes.cC`s |dkS(Ni((R((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pyRss -dumpversiont GCC_NOT_FOUNDcC`stjddS(s Return 32 for 32bit arch, 64 for 64bit arch. By "architecture", we mean the size of memory pointers (size_t in C), *not* the size of long int, as it can be different. tPi(tstructtcalcsize(((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytlocal_bitwidth,s cC`stjddS(s Return the bit width of Python int (C long int). Note that it can be different from the size of a memory pointer. Ri(R+R,(((s5/tmp/pip-build-X4mzal/theano/theano/configdefaults.pytpython_int_bitwidth;sRt processortpython_versiontpython_bitwidthR.ttheano_versiont numpy_versionRWt_t gxx_versionthostnamecC`s|dkrtj}n|dkr6tj}n|jd}t|dkr[|S|djd}t|dkr|Sdj|d |d<|djd}x+t|r|djr|d=qPqWdj||ds            @                         '      -                                                                    !                          K      + $!