2026-07-18:15:17:55 WARNING [config.evaluate_config:281] --limit SHOULD ONLY BE USED FOR TESTING. REAL METRICS SHOULD NOT BE COMPUTED USING LIMIT. 2026-07-18:15:17:58 INFO [_cli.run:376] Selected Tasks: ['babilong_longctx'] 2026-07-18:15:18:00 INFO [evaluator:211] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 2026-07-18:15:18:00 INFO [evaluator:236] Initializing hf model, with arguments: {'pretrained': '/workspace/outputs/l2a_style/stage2/checkpoint-25', 'trust_remote_code': True, 'dtype': 'bfloat16', 'max_length': 16384, 'attn_implementation': 'sdpa'} 2026-07-18:15:18:03 INFO [models.huggingface:161] Using device 'cuda:0' 2026-07-18:15:18:03 INFO [models.huggingface:548] Model type cannot be determined. Using default model type 'causal' 2026-07-18:15:18:03 INFO [models.huggingface:423] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} 2026-07-18:15:18:05 WARNING [api.task:856] babilong_qa5: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager. For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme. 2026-07-18:15:18:05 WARNING [api.task:856] babilong_qa4: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager. For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme. 2026-07-18:15:18:05 WARNING [api.task:856] babilong_qa3: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager. For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme. 2026-07-18:15:18:06 WARNING [api.task:856] babilong_qa2: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager. For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme. 2026-07-18:15:18:06 WARNING [api.task:856] babilong_qa1: Custom kwargs can be passed to `--metadata` in console (as json string) or to the TaskManager. For example --metadata='{"max_seq_lengths":[4096, 8192]}'. For details see task Readme. 2026-07-18:15:18:06 INFO [tasks:700] Selected tasks: 2026-07-18:15:18:06 INFO [tasks:703] Group: babilong_longctx 2026-07-18:15:18:06 INFO [tasks:726] ConfigurableGroup(group=babilong_longctx,group_alias=None): {'babilong_qa1': ConfigurableTask(task_name=babilong_qa1,output_type=generate_until,num_fewshot=2,num_samples=1000), 'babilong_qa2': ConfigurableTask(task_name=babilong_qa2,output_type=generate_until,num_fewshot=2,num_samples=999), 'babilong_qa3': ConfigurableTask(task_name=babilong_qa3,output_type=generate_until,num_fewshot=2,num_samples=999), 'babilong_qa4': ConfigurableTask(task_name=babilong_qa4,output_type=generate_until,num_fewshot=2,num_samples=999), 'babilong_qa5': ConfigurableTask(task_name=babilong_qa5,output_type=generate_until,num_fewshot=2,num_samples=999)} 2026-07-18:15:18:06 INFO [evaluator:314] babilong_qa1: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 16, 'until': []} 2026-07-18:15:18:06 WARNING [evaluator:333] Overwriting default num_fewshot of babilong_qa1 from 2 to 2 2026-07-18:15:18:06 INFO [evaluator:314] babilong_qa2: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 16, 'until': []} 2026-07-18:15:18:06 WARNING [evaluator:333] Overwriting default num_fewshot of babilong_qa2 from 2 to 2 2026-07-18:15:18:06 INFO [evaluator:314] babilong_qa3: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 16, 'until': []} 2026-07-18:15:18:06 WARNING [evaluator:333] Overwriting default num_fewshot of babilong_qa3 from 2 to 2 2026-07-18:15:18:06 INFO [evaluator:314] babilong_qa4: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 16, 'until': []} 2026-07-18:15:18:06 WARNING [evaluator:333] Overwriting default num_fewshot of babilong_qa4 from 2 to 2 2026-07-18:15:18:06 INFO [evaluator:314] babilong_qa5: Using gen_kwargs: {'do_sample': False, 'temperature': 0.0, 'max_gen_toks': 16, 'until': []} 2026-07-18:15:18:06 WARNING [evaluator:333] Overwriting default num_fewshot of babilong_qa5 from 2 to 2 2026-07-18:15:18:06 INFO [api.task:311] Building contexts for babilong_qa1 on rank 0... 0%| | 0/50 [00:00