4b284b21bc4/eval/agg.4b284b21bc4_GEM-web_nlg_en_implicit-graph-description_2.json

{"results": [{"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "bleu": 1.8053876932339799, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "bleu_stderr": 0.0692633939605246}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge1_precision": 0.0953355786305315, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge1_precision_stderr": 0.0018482318125930552}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge1_recall": 0.587435498341808, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge1_recall_stderr": 0.0044160353671122265}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge1_fmeasure": 0.15238609977285375, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge1_fmeasure_stderr": 0.0021050373299750613}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge2_precision": 0.04635353043989582, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge2_precision_stderr": 0.0011008091437590098}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge2_recall": 0.31351855748984986, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge2_recall_stderr": 0.0042413072919440875}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rouge2_fmeasure": 0.07451843059126674, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rouge2_fmeasure_stderr": 0.0013538507376262556}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeL_precision": 0.08027003022288653, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeL_precision_stderr": 0.0014089308706206984}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeL_recall": 0.5250978321957749, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeL_recall_stderr": 0.0044081811988050555}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeL_fmeasure": 0.12964573051803532, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeL_fmeasure_stderr": 0.0015563770094126131}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeLsum_precision": 0.08462110334448762, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeLsum_precision_stderr": 0.0017266673212118049}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeLsum_recall": 0.5256991157959195, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeLsum_recall_stderr": 0.004362166200252256}, {"task_name": "GEM/web_nlg_en", "prompt_name": "implicit-graph-description", "rougeLsum_fmeasure": 0.13493557268343806, "fixed_answer_choice_list": null, "dataset_path": "GEM/web_nlg", "dataset_name": "en", "subset": null, "prompt_id": "38342608-5cd7-4ce7-b2e1-905ecd7f4c80", "prompt_jinja": "{{input | join(\"; \")}}\nThe above is a collection of relations. Write descriptive English that contains this information.\n{% for i in references %}\n  ||| {{ i }} \n{% endfor %}", "prompt_original_task": true, "comment": "", "rougeLsum_fmeasure_stderr": 0.0019537513800864143}], "config": {"model": "hf-causal", "model_args": "pretrained=/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/lm1-4b2-84b-c4-repetitions/4b284b21bc4/transformers,use_accelerate=True,tokenizer=/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/gpt2,dtype=bfloat16", "task_args": "", "num_fewshot": 2, "batch_size": 16, "device": "cuda", "use_cache": false, "limit": 3000, "bootstrap_iters": 10, "seed": 1234}}