import json import os import requests import pandas as pd dataset_link = "[`tweetner7`](https://huggingface.co/datasets/tner/tweetner7)" metric_dir = 'metric_files' os.makedirs(metric_dir, exist_ok=True) def lm_link(_model): return f"[`{_model}`](https://huggingface.co/{_model})" def model_link(_model, _type): return f"[`tner/{_model}-tweetner7-{_type}`](https://huggingface.co/tner/{_model}-tweetner7-{_type})" def download(_model, _type): url = f"https://huggingface.co/tner/{_model}-tweetner7-{_type}/raw/main/eval" filename = f"{metric_dir}/{_model}-{_type}.json" print(url, filename) try: with open(filename) as f: return json.load(f) except Exception: tmp = {} for metric in ["metric.test_2021", "metric.test_2020", "metric_span.test_2021", "metric_span.test_2020"]: year = metric[-4:] if metric not in tmp: _metric = json.loads(requests.get(f"{url}/{metric}.json").content) if '_span' in metric: tmp[f"Entity-Span F1 ({year})"] = round(100 * _metric["micro/f1"], 2) else: tmp[f"Micro F1 ({year})"] = round(100 * _metric["micro/f1"], 2) tmp[f"Macro F1 ({year})"] = round(100 * _metric["macro/f1"], 2) tmp.update({f"F1 ({year})/{k}": round(100 * v['f1'], 2) for k, v in _metric["per_entity_metric"].items()}) with open(filename, "w") as f: json.dump(tmp, f) return tmp lms = [ "roberta-large", "roberta-base", "cardiffnlp/twitter-roberta-base-2019-90m", "cardiffnlp/twitter-roberta-base-dec2020", "cardiffnlp/twitter-roberta-base-dec2021" "vinai/bertweet-large", "vinai/bertweet-base", "bert-large", "bert-base" ] types = [ ["all", "continuous", "2021", "2020"], ["random"], [ "selflabel2020", "selflabel2021", "2020-selflabel2020-all", "2020-selflabel2021-all", "selflabel2020-continuous", "selflabel2021-continuous" ] ] for tt in types: metrics = [] for t in tt: for lm in lms: if 'selflabel' in t and lm != "roberta-large": continue _lm_link = lm_link(lm) lm = os.path.basename(lm) _model_link = model_link(lm, t) __metric = { "Model (link)": model_link(lm, t), "Data": dataset_link, "Language Model": _lm_link } __metric.update(download(lm, t)) metrics.append(__metric) df = pd.DataFrame(metrics) print(tt) print(df.to_markdown(index=False)) print()