Support base_workload=all to run all workload types

[maryamtahhan]

Enhancement Request

Currently, the llm-benchmark-concurrent-load.yml playbook requires specifying a single base_workload parameter:

ansible-playbook -i inventory/hosts.yml \
  llm-benchmark-concurrent-load.yml \
  -e "test_model=meta-llama/Llama-3.2-1B-Instruct" \
  -e "base_workload=chat" \
  -e "requested_cores=16"

Proposed Enhancement

Support base_workload=all to automatically loop over all available workload types:

ansible-playbook -i inventory/hosts.yml \
  llm-benchmark-concurrent-load.yml \
  -e "test_model=meta-llama/Llama-3.2-1B-Instruct" \
  -e "base_workload=all" \
  -e "requested_cores=16"

This would automatically run all workloads:

• chat - Chat workload (512:256)
• rag - RAG workload (4096:512)
• code - Code generation (512:4096)
• summarization - Summarization (1024:256)
• short_codegen - Short code generation

Benefits

1. Comprehensive testing: Easy way to run all workload types for a model
2. Reduced manual intervention: No need to run the playbook multiple times
3. Better coverage: Ensures all workload types are tested consistently

Implementation Notes

The playbook would need to:

1. Detect when base_workload=all is specified
2. Loop over all available workload types
3. Run the 3-phase testing for each workload
4. Collect and organize results by workload type

Related file: automation/test-execution/ansible/llm-benchmark-concurrent-load.yml