from opencompass.models.turbomind import TurboMindModel models = [ dict( type=TurboMindModel, abbr='internlm2-20b-turbomind', path='internlm/internlm2-20b', engine_config=dict( session_len=32768, max_batch_size=32, model_name='internlm2-20b', tp=2, ), gen_config=dict( top_k=1, top_p=0.8, temperature=1.0, max_new_tokens=2000, ), max_out_len=2000, max_seq_len=32768, batch_size=32, concurrency=8, run_cfg=dict(num_gpus=2, num_procs=1), ) ]