clarify cache folder

Author: Jeronymous

whisper_timestamped/transcribe.py

@@ -2392,7 +2392,7 @@ def load_model(
     backend : str, optional
         Backend to use. Either "transformers" or "openai-whisper".
     download_root : str, optional
-        Root folder to download the model to. If None, use the default download root.
+        Root folder to download the model to. If None, use the default download root (typically: ~/.cache)
     in_memory : bool, optional
         Whether to preload the model weights into host memory.
     """
@@ -2405,11 +2405,12 @@ def load_model(
             name = f"openai/whisper-{name}"
         # TODO: use download_root
         # TODO: does in_memory makes sense?
+        cache_dir=os.path.join(download_root, "huggingface", "hub") if download_root else None,
         try:
-            generation_config = transformers.GenerationConfig.from_pretrained(name)
+            generation_config = transformers.GenerationConfig.from_pretrained(name, cache_dir=cache_dir)
         except OSError:
-            generation_config = transformers.GenerationConfig.from_pretrained("openai/whisper-tiny")
-        processor = transformers.WhisperProcessor.from_pretrained(name)
+            generation_config = transformers.GenerationConfig.from_pretrained("openai/whisper-tiny", cache_dir=cache_dir)
+        processor = transformers.WhisperProcessor.from_pretrained(name, cache_dir=cache_dir)
         if device is None:
             device = "cuda" if torch.cuda.is_available() else "cpu"
         precision = torch.float32
@@ -2421,6 +2422,7 @@ def load_model(
             # torch_dtype=torch.bfloat16, 
             # attn_implementation="flash_attention_2",
             # attn_implementation="sdpa",
+            cache_dir=cache_dir,
         )
         # model = model.to_bettertransformer()
 
@@ -2433,7 +2435,12 @@ def load_model(
     extension = os.path.splitext(name)[-1] if os.path.isfile(name) else None
 
     if name in whisper.available_models() or extension == ".pt":
-        return whisper.load_model(name, device=device, download_root=download_root, in_memory=in_memory)
+        return whisper.load_model(
+            name,
+            device=device,
+            download_root=os.path.join(download_root, "whisper") if download_root else None,
+            in_memory=in_memory
+        )
 
     # Otherwise, assume transformers
     if extension in [".ckpt", ".bin"]:
@@ -2446,7 +2453,11 @@ def load_model(
             raise ImportError(f"If you are trying to download a HuggingFace model with {name}, please install first the transformers library")
         from transformers.utils import cached_file
 
-        kwargs = dict(cache_dir=download_root, use_auth_token=None, revision=None)
+        kwargs = dict(
+            cache_dir=os.path.join(download_root, "huggingface", "hub") if download_root else None,
+            use_auth_token=None,
+            revision=None,
+        )
         try:
             model_path = cached_file(name, "pytorch_model.bin", **kwargs)
         except OSError as err: