from opencompass.models import HuggingFaceCausalLM models = [ # WizardCoder Python 34B dict( type=HuggingFaceCausalLM, abbr='WizardCoder-Python-34B-V1.0', path="WizardLM/WizardCoder-Python-34B-V1.0", tokenizer_path='WizardLM/WizardCoder-Python-34B-V1.0', tokenizer_kwargs=dict( padding_side='left', truncation_side='left', trust_remote_code=True, ), max_out_len=1024, max_seq_len=2048, batch_size=8, model_kwargs=dict(trust_remote_code=True, device_map='auto'), run_cfg=dict(num_gpus=4, num_procs=1), ), ]