[Bug] Make free_cache_engine option workable in latest vLLM/SGLang
#1414
Annotations
1 error
|
Running GSM8K E2E training tests with 3D parallelism on 8 L20 GPUs with tie-embedding Megatron (Qwen) with train tp > infer tp
Process completed with exit code 1.
|
Loading