cache dir

app.py

@@ -1,6 +1,6 @@
 import evaluate
 from evaluate.utils import launch_gradio_widget
 
-CACHE_DIR="/gfs/team/nlp/users/rcardena/tools/huggingface/evaluate"
-module = evaluate.load("my_perplexity", module_type="measurement",cache_dir=CACHE_DIR)
+# CACHE_DIR="/gfs/team/nlp/users/rcardena/tools/huggingface/evaluate"
+module = evaluate.load("my_perplexity", module_type="measurement")#,cache_dir=CACHE_DIR)
 launch_gradio_widget(module)

my_perplexity.py

@@ -106,7 +106,7 @@ class MyPerplexity(evaluate.Metric):
 
     def _compute(self, predictions, model_id, batch_size: int = 16, add_start_token: bool = True, device=None):
 
-        CACHE_DIR="/gfs/team/nlp/users/rcardena/tools/huggingface/evaluate"
+        # CACHE_DIR="/gfs/team/nlp/users/rcardena/tools/huggingface/evaluate"
         if device is not None:
             assert device in ["gpu", "cpu", "cuda"], "device should be either gpu or cpu."
             if device == "gpu":
@@ -114,12 +114,12 @@ class MyPerplexity(evaluate.Metric):
         else:
             device = "cuda" if torch.cuda.is_available() else "cpu"
 
-        model = AutoModelForCausalLM.from_pretrained(model_id,cache_dir=CACHE_DIR)
+        model = AutoModelForCausalLM.from_pretrained(model_id)#,cache_dir=CACHE_DIR)
         model = model.to(device)
 
         tokenizer = AutoTokenizer.from_pretrained(
                             model_id,
-                            cache_dir=CACHE_DIR,
+                            # cache_dir=CACHE_DIR,
                             use_fast="cnn_dailymail" not in model_id,
                             )