:j"ddlZddlZddlZddlmZddlmZmZddlZddl m Z ddl m Z ddl mZmZmZddlmZmZddlmZdd lmZdd lmZmZdd lmZdd lmZej>e Z!e d Z"e dZ#dZ$de%e&fdZ'e"r ddl(Z(e(jRdGdZ*e*jWdddie(jXde(jZde(j\de(j^e$dde(j\de(j^e'dde(j\de&d d!e(j\d"d#e0d$d%e(j\d&d'e0dd(e(j\d)d*e&d+d,e(j\d-d.e0d/d0e(j\d1e(j^e ee ezdd2e(jXd3d4e(jb5d6eddfd7Z2e*jWd8ddie(jXd9e(jZde(j\d:e(jZddd;d9e&deddfd?Z4d@e&dAe&de0fdBZ5d>ed3e%e&ddfdCZ6dHd>ed3ee%e&ddfdDZ7e dEk(r#e"se!jqdFe9d$e2yy)IN) Namespace)AnyOptional)RequirementCache)get_args)CPUAcceleratorCUDAAcceleratorMPSAccelerator)_PRECISION_INPUT_STR_PRECISION_INPUT_STR_ALIAS)STRATEGY_REGISTRY)_process_cli_args)_parse_gpu_ids_select_auto_accelerator)_suggested_max_num_threads)_load_distributed_checkpointclick lightning_sdk)cpugpucudampstpuautoreturnctj}d}|Dcgc]}tj||r|c}Scc}w)zReturns strategy choices from the registry, with the ones removed that are incompatible to be launched from the CLI or ones that require further configuration by the user.z).*(spawn|fork|notebook|xla|tpu|offload).*)r available_strategiesrematch)rexcludedstrategys i/home/mcse/projects/srt_converter/srt-converter-venv/lib/python3.12/site-packages/lightning/fabric/cli.py_get_supported_strategiesr#(s:-AAC;H%9 ^(T\A]H ^^ ^s>>cyNr&r"_mainr(3s r'runignore_unknown_optionsT)context_settingsscript)exists)typez --acceleratorz#The hardware accelerator to run on.)r.defaulthelpz --strategyz0Strategy for how to run across multiple devices.z --devices1zNumber of devices to run on (``int``), which devices to run on (``list`` or ``str``), or ``'auto'``. The value applies per node.z --num-nodesz --num_nodesz5Number of machines (nodes) for distributed execution.z --node-rankz --node_rankzrThe index of the machine (node) this command gets started on. Must be a number in the range 0, ..., num_nodes - 1.z--main-addressz--main_addressz 127.0.0.1zTThe hostname or IP address of the main machine (usually the one with node_rank = 0).z --main-portz --main_portirz-The main port to connect to the main machine.z --precisionzDouble precision (``64-true`` or ``64``), full precision (``32-true`` or ``32``), half precision (``16-mixed`` or ``16``) or bfloat16 precision (``bf16-mixed`` or ``bf16``) script_args)nargsr.kwargsc ft|jdg}ttdi||y)aRun a Lightning Fabric script. SCRIPT is the path to the Python script with the code to run. The script must contain a Fabric object. SCRIPT_ARGS are the remaining arguments that you can pass to the script itself and are expected to be parsed there. r3argsr3Nr&)listpopmainr)r6r3s r"_runr=7s,b6::mR89  )%f%;?r' consolidatecheckpoint_folderz --output_filezPath to the file where the converted checkpoint should be saved. The file should not already exist. If no path is provided, the file will be saved next to the input checkpoint folder with the same name and a '.consolidated' suffix. output_filect||}t|}t|j}t j ||j y)zConvert a distributed/sharded checkpoint into a single file that can be loaded with `torch.load()`. Only supports FSDP sharded checkpoints at the moment. )r?r@N)rrrr?torchsaver@)r?r@r9config checkpoints r" _consolidaterFs?4+<+V"4(1&2J2JK  :v112r'r9cdtjd<|j&t|jtjd<|j&t|jtjd<t|j tjd<t|j tjd<|j't|jtjd<yy) zwSet the environment variables for the new processes. The Fabric connector will parse the arguments set here. r1 LT_CLI_USEDNLT_ACCELERATOR LT_STRATEGY LT_DEVICES LT_NUM_NODES LT_PRECISION)osenviron acceleratorstrr!devices num_nodes precision)r9s r"_set_env_variablesrUs !$BJJ} #'*4+;+;'< #$ }} $' $6 =!"4<<0BJJ|!$T^^!4BJJ~ ~~!%(%8 >""r'rPrRcl|dk(s| t}|dk(r!|dk(s |dk(s|dk(rd}ntd|d|dk(rt|d d }n[|dk(rtj|}n@|dk(rt j|}n%|d k(r td t j|S| t|Sd S)zhParse the `devices` argument to determine how many processes need to be launched on the current machine.rrrrr1z.Cannot default to '1' device for accelerator=''rT) include_cuda include_mpsrz=Launching processes for TPU through the CLI is not supported.r)r ValueErrorrr parse_devicesr rlen)rPrRparsed_devicess r"_get_num_processesr^sf 3.0 & & K5$8K5  XYY++G44"0"<3~ C!Cr'cddlmcm}|jdk(rdnt |j |j }d|d|jd|jd|jd |j|jg}|j|tjjd t!t#|j%|y) zYThis will invoke `torchrun` programmatically to launch the given script in new processes.rNdpr2z--nproc_per_node=z --nnodes=z --node_rank=z--master_addr=z--master_port=OMP_NUM_THREADS)torch.distributed.run distributedr)r!r^rPrRrS node_rank main_address main_portr,extendrNrO setdefaultrQrr<)r9r3torchrun num_processes torchrun_argss r"_torchrun_launchrls,,$.A4FtGWGWY]YeYe4fM M?+ DNN#$ t~~&' **+, () M%JJ+S1K1M-NO MM- r'c:t|t||xsgyr%)rUrlr8s r"r<r<stT;,"-r'__main__zoTo use the Lightning Fabric CLI, you must have `click` installed. Install it by running `pip install -U click`.)rNr%):loggingrNrargparsertypingrrrB lightning_utilities.core.importsrtyping_extensionsrlightning.fabric.acceleratorsrr r ,lightning.fabric.plugins.precision.precisionr r lightning.fabric.strategiesr 1lightning.fabric.utilities.consolidate_checkpointr(lightning.fabric.utilities.device_parserrr&lightning.fabric.utilities.distributedrlightning.fabric.utilities.loadr getLogger__name___log_CLICK_AVAILABLE_LIGHTNING_SDK_AVAILABLE_SUPPORTED_ACCELERATORSr:rQr#rgroupr(commandargumentPathoptionChoiceint UNPROCESSEDr=rFrUr^rlr<error SystemExitr&r'r"rs  =&YYi9O]MHw"#G,+O<F_49_U[[]   ]] $d  U^^ UZZt $U\\ U\\1 2 2  U\\ U\\35 6 ?  U\\  + U\\  D U\\  & U\\  c U\\  < U\\ U\\(#788D^;__ ` i U^^M%2C2CD @s @t @E     P @ ]] $d  U^^ UZZt $U\\ UZZt $ -  3 3(3- 3D 3   ( 39Y949"DCD#D#D.!9!49!!*.y.xS ':.d.  z   = mFr'