o TiC@s>ddlZddlZddlZddlZddlZddlmZddlmZm Z m Z ddl m Z dZ dZdZd Ze d eZe d eZed eZed eZdZee eeZe d eZed eZd(ddZddZddZddZddZddZddZ ddZ!d d!Z"d)d#d$Z#d%d&Z$e%d'kre#dSdS)*N)ALL_OPS) installed_ops torch_infoaccelerator_name)get_acceleratorzzzzz [SUCCESS] z[OKAY]z [WARNING]z[FAIL]z[INFO]Tc Cs&d}d}gd}td||t|dt|dtdtd||t|dt|dtdtd||t|dt|dtd trUtnt}td d |d |td||t|dt|dt|dd |t|d|dd |t|d|d td||t|dt|dtdt}tdt}t D]C\}}d |t|} | |rtn|} t |drt tjkr|n|} d t|d|t|dt| t} t|| | | | qtd||t|dt|ddS)N )zop name installed compatible-rrz&DeepSpeed C++/CUDA extension op reportzNOTE: Ops not installed will be just-in-time (JIT) compiled at runtime if needed. Op compatibility means that your system meet the required dependencies to JIT install the op.zJIT compiled ops requires ninjaninja.z[YES]z[NO]F)printlenninja_installedOKAYFAILGREENENDYELLOWritems is_compatiblergetrr_name color_len) verbosemax_dots max_dots2h ninja_statusr noop_namebuilderdotsr is_installeddots2r)H/home/ubuntu/.local/lib/python3.10/site-packages/deepspeed/env_report.py op_reports:((((<( ,,r+cCs$zddl}WdStyYdSw)NrFT)r ImportError)r r)r)r*r=s   rcCsddl}|jjj}|durtd|jjjdtSz tj|ddgdd}Wnty7tdtYSw| }| d }||d  d d  d }d |S)NrzF [FAIL] cannot find CUDA_HOME via torch.utils.cpp_extension.CUDA_HOME= z /bin/nvccz-VT)universal_newlinesz [FAIL] nvcc missing releaser,r) torch.utils.cpp_extensionutils cpp_extension CUDA_HOMEREDr subprocess check_outputFileNotFoundErrorsplitindexreplacejoin)torch cuda_homeoutput output_split release_idxr/r)r)r* nvcc_versionEs    rCcCs*dtjvstjtjdrtjdSdS)NASCEND_HOME_PATH)osenvironpathexistsr)r)r)r*installed_cann_pathTs rIc sddlt}|dur dSd}ttj|D]J\}}}|r"|Sfdd|D}|rbtj||d}t|d}|D]}|ddkrR| d d}nq>Wdn1s]wYq|S) Nrz5CANN_HOME does not exist, unable to compile NPU op(s)r1csg|] }d|r|qS)zascend_.*_install\.info)match).0filerer)r* csz*installed_cann_version..rversion=) rNrIrEwalkrGrealpathr=openfindstripr:) ascend_path cann_versiondirpath_ filenames install_filesfilepathfliner)rMr*installed_cann_versionZs,  rbc Csztd}WnttttfyYdSw|j|j}t|}g}|dkrC| dt dt t dkrC| dt dt ||fS)Nz/dev/shm)UNKNOWNNi r-zh [WARNING] /dev/shm size might be too small, if running in docker increase to at least --shm-size='1gb' ncclz [WARNING] see more details about NCCL requirements: https://docs.nvidia.com/deeplearning/nccl/user-guide/docs/troubleshooting.html#sharing-data )rEstatvfsOSErrorr9 ValueErrorAttributeErrorf_frsizef_blockshuman_readable_sizeappendrrrcommunication_backend_name) shm_statsshm_size shm_hbyteswarnr)r)r* get_shm_sizens" rrcCs`gd}d}|dkr&|t|dkr&|d}|d7}|dkr&|t|dks|dd||S)N)BKBMBGBTBrirz.2fr-)r)sizeunitsir)r)r*rksrkc Csd}dtjfdtjfdtjfdtjdtjdtjfg}tdkr[ttj dd}| d tj j fd |fd |r;dnt fd d t dd|rOdt dndt dfgn3tdkrddl}| d d t dfd|jfd|jfdtfgn| d d t ddfg|dtftd|D]&\}}g}t|tr|\}}t|d|t|||r|D]}t|qqdS)N!ztorch install pathz torch versionzdeepspeed install pathzdeepspeed infoz, cudahipztorch cuda versionztorch hip versionz nvcc versionzdeepspeed wheel compiled w.ztorch rQzhip hip_versionzcuda cuda_versionnpurztorch_npu install pathztorch_npu versionzascend_cann versionr-zshared memory (/dev/shm) sizez#DeepSpeed general environment info:r)r>__path__ __version__ deepspeed __git_hash____git_branch__r device_namegetattrrQextendr|rCr torch_npurbrlrrr isinstancetupler)rreportr~rnamevaluewarnsrqr)r)r* debug_reports@   rcCs4t}|jdddd|jdddd|}|S)Nz--hide_operator_status store_truezTSuppress display of installation and compatibility statuses of DeepSpeed operators. )actionhelpz--hide_errors_and_warningsz$Suppress warning and error messages.)argparseArgumentParser add_argument parse_args)parserargsr)r)r*parse_argumentssrFcCs|st| dtdS)N)r)r+rhide_operator_statushide_errors_and_warningsr)r)r*mains  rcCst}t|j|jddS)Nr)rrrr)rr)r)r*cli_mainsr__main__)T)FF)&rEr>rr7rops.op_builder.all_opsrgit_version_inforrrdeepspeed.acceleratorrrr6rrSUCCESSrWARNINGrINFOrrokaywarningr+rrCrIrbrrrkrrrr__name__r)r)r)r*sD    "