Make sure cache is skipped for HF datasets

data.py

@@ -60,12 +60,12 @@ def read_one_dataframe(json_path: str, device_label: str) -> pd.DataFrame:
 def get_distant_data() -> pd.DataFrame:
     # Retrieve AMD dataframe
     amd_src = "hf://datasets/optimum-amd/transformers_daily_ci/**/runs/**/ci_results_run_models_gpu/model_results.json"
-    files_amd = sorted(fs.glob(amd_src), reverse=True)
+    files_amd = sorted(fs.glob(amd_src, refresh=True), reverse=True)
     df_amd = read_one_dataframe(f"hf://{files_amd[0]}", "amd")
     # Retrieve NVIDIA dataframe, which pattern should be:
     # hf://datasets/hf-internal-testing/transformers_daily_ci/raw/main/YYYY-MM-DD/ci_results_run_models_gpu/model_results.json 
     nvidia_src = "hf://datasets/hf-internal-testing/transformers_daily_ci/*/ci_results_run_models_gpu/model_results.json"
-    files_nvidia = sorted(fs.glob(nvidia_src), reverse=True)
+    files_nvidia = sorted(fs.glob(nvidia_src, refresh=True), reverse=True)
     # NOTE: should this be removeprefix instead of lstrip?
     nvidia_path = files_nvidia[0].lstrip('datasets/hf-internal-testing/transformers_daily_ci/')
     nvidia_path = "https://huggingface.co/datasets/hf-internal-testing/transformers_daily_ci/raw/main/" + nvidia_path