Fix: Support custom HF_ENDPOINT for model recipe downloads (#1982)

SUMMARY:
Replace hardcoded `HUGGINGFACE_CO_URL_HOME` with `HfApi().`endpoint to
enable custom HuggingFace endpoint configuration via `HF_ENDPOINT`
environment variable. (Fixed #1967)


TEST PLAN:
Tested with custom endpoint (`HF_ENDPOINT=https://hf-mirror.com`): Uses
mirror site.

Any feedback or suggestions for improvement are welcome!

---------

Signed-off-by: Shuting Zhao <shuting.zhao@outlook.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: HDCharles <39544797+HDCharles@users.noreply.github.com>

Author: 3 people

src/llmcompressor/transformers/utils/helpers.py

@@ -10,7 +10,7 @@
 import requests
 from huggingface_hub import (
     _CACHED_NO_EXIST,
-    HUGGINGFACE_CO_URL_HOME,
+    HfApi,
     hf_hub_download,
     try_to_load_from_cache,
 )
@@ -149,7 +149,9 @@ def recipe_from_huggingface_model_id(
         logger.debug("HF_HUB_OFFLINE is set, skipping recipe download from HuggingFace")
         return None
 
-    model_id_url = os.path.join(HUGGINGFACE_CO_URL_HOME, hf_stub)
+    # Use custom HF_ENDPOINT
+    hf_api = HfApi()
+    model_id_url = f"{hf_api.endpoint.rstrip('/')}/{hf_stub}"
     request = requests.head(model_id_url)
 
     if request.status_code != 200: