This repository was archived by the owner on Sep 24, 2025. It is now read-only.
File tree Expand file tree Collapse file tree
Expand file tree Collapse file tree Original file line number Diff line number Diff line change @@ -277,6 +277,11 @@ def ilab_pipeline(
277277 data_processing_task .after (model_to_pvc_task , sdg_task )
278278 data_processing_task .set_caching_options (False )
279279 data_processing_task .set_env_variable ("XDG_CACHE_HOME" , "/tmp" )
280+ data_processing_task .set_accelerator_type (eval_gpu_identifier )
281+ data_processing_task .set_accelerator_limit (1 )
282+ add_toleration_json (data_processing_task , train_tolerations )
283+ add_node_selector_json (data_processing_task , train_node_selectors )
284+
280285
281286 # Upload "skills_processed_data" and "knowledge_processed_data" artifacts to S3 without blocking the rest of the workflow
282287 skills_processed_data_to_artifact_task = skills_processed_data_to_artifact_op ()
Original file line number Diff line number Diff line change @@ -934,6 +934,11 @@ deploymentSpec:
934934 - name : XDG_CACHE_HOME
935935 value : /tmp
936936 image : registry.redhat.io/rhelai1/instructlab-nvidia-rhel9@sha256:3e6eb035c69b204746a44b3a58b2751c20050cfb6af2ba7989ba327809f87c0b
937+ resources :
938+ accelerator :
939+ count : ' 1'
940+ resourceCount : ' 1'
941+ resourceType : ' {{$.inputs.parameters['' pipelinechannel--eval_gpu_identifier'' ]}}'
937942 exec-deletepvc :
938943 container :
939944 image : argostub/deletepvc
@@ -2620,8 +2625,13 @@ root:
26202625 - sdg-op
26212626 inputs :
26222627 parameters :
2628+ accelerator_type :
2629+ runtimeValue :
2630+ constant : ' {{$.inputs.parameters['' pipelinechannel--eval_gpu_identifier'' ]}}'
26232631 max_batch_len :
26242632 componentInputParameter : sdg_max_batch_len
2633+ pipelinechannel--eval_gpu_identifier :
2634+ componentInputParameter : eval_gpu_identifier
26252635 taskInfo :
26262636 name : data-processing-op
26272637 deletepvc :
@@ -3395,6 +3405,9 @@ platforms:
33953405 deploymentSpec :
33963406 executors :
33973407 exec-data-processing-op :
3408+ nodeSelector :
3409+ nodeSelectorJson :
3410+ componentInputParameter : train_node_selectors
33983411 pvcMount :
33993412 - mountPath : /model
34003413 pvcNameParameter :
@@ -3412,6 +3425,9 @@ platforms:
34123425 taskOutputParameter :
34133426 outputParameterKey : name
34143427 producerTask : createpvc
3428+ tolerations :
3429+ - tolerationJson :
3430+ componentInputParameter : train_tolerations
34153431 exec-generate-metrics-report-op :
34163432 pvcMount :
34173433 - mountPath : /output
You can’t perform that action at this time.
0 commit comments