b1ac8dc1bc
* BF16 Lora Finetuning on K8S with OneCCL and Intel MPI * Update README.md * format * refine * Update README.md * refine * Update README.md * increase nfs volume size to improve IO performance * fix bugs * Update README.md * Update README.md * fix permission * move output destination * Update README.md * fix wrong base model name in doc * fix output path in entrypoint * add a permission-precreated output dir * format * move output logs to a persistent storage
11 lines
617 B
YAML
11 lines
617 B
YAML
imageName: intelanalytics/bigdl-lora-finetuning:2.4.0-SNAPSHOT
|
|
trainerNum: 8
|
|
microBatchSize: 8
|
|
TEEMode: tdx # tdx or native
|
|
nfsServerIp: your_nfs_server_ip
|
|
nfsPath: a_nfs_shared_folder_path_on_the_server
|
|
dataSubPath: alpaca_data_cleaned_archive.json # a subpath of the data file under nfs directory
|
|
modelSubPath: llama-7b-hf # a subpath of the model file (dir) under nfs directory
|
|
outputSubPath: output # a subpath of the empty directory under the nfs directory to save finetuned model, for example, if you make an empty dir named 'output' at the nfsPath, the value should be 'output'
|
|
ompNumThreads: 14
|
|
cpuPerPod: 42
|