chore: move clip models

client/setup.py

@@ -43,7 +43,7 @@
     setup_requires=['setuptools>=18.0', 'wheel'],
     install_requires=[
         'jina>=3.12.0',
-        'docarray[common]>=0.19.0,<0.30.0',
+        'docarray[common]==0.21.0',
         'packaging',
     ],
     extras_require={

server/clip_server/model/clip_onnx.py

@@ -1,5 +1,6 @@
 import os
 from typing import Dict, Optional
+import requests
 
 from clip_server.model.pretrained_models import (
     download_model,
@@ -8,10 +9,7 @@
 )
 from clip_server.model.clip_model import BaseCLIPModel
 
-_S3_BUCKET = (
-    'https://clip-as-service.s3.us-east-2.amazonaws.com/models/onnx/'  # Deprecated
-)
-_S3_BUCKET_V2 = 'https://clip-as-service.s3.us-east-2.amazonaws.com/models-436c69702d61732d53657276696365/onnx/'
+_HUGGINGFACE_ONNX_BUCKET = 'https://huggingface.co/jinaai/clip-models/'
 _MODELS = {
     'RN50::openai': (
         ('RN50/textual.onnx', '722418bfe47a1f5c79d1f44884bb3103'),
@@ -213,14 +211,14 @@ def __init__(
                 )
                 textual_model_name, textual_model_md5 = _MODELS[name][0]
                 self._textual_path = download_model(
-                    url=_S3_BUCKET_V2 + textual_model_name,
+                    url=self.get_onnx_model_url(name=textual_model_name),
                     target_folder=cache_dir,
                     md5sum=textual_model_md5,
                     with_resume=True,
                 )
                 visual_model_name, visual_model_md5 = _MODELS[name][1]
                 self._visual_path = download_model(
-                    url=_S3_BUCKET_V2 + visual_model_name,
+                    url=self.get_onnx_model_url(name=visual_model_name),
                     target_folder=cache_dir,
                     md5sum=visual_model_md5,
                     with_resume=True,
@@ -261,6 +259,23 @@ def get_model_name(name: str):
 
         return name
 
+    @staticmethod
+    def get_onnx_model_url(name: str):
+        hf_download_url = (
+            _HUGGINGFACE_ONNX_BUCKET
+            + 'resolve/main/'
+            + name.split('/')[0]
+            + '-'
+            + name.split('/')[1]
+            + '?download=true'
+        )
+        try:
+            response = requests.head(hf_download_url, timeout=10)
+            if response.status_code in [200, 302]:
+                return hf_download_url
+        except Exception:
+            raise ValueError('Invalid model url.')
+
     def start_sessions(
         self,
         dtype,

server/clip_server/model/pretrained_models.py

@@ -2,9 +2,10 @@
 import hashlib
 import shutil
 import urllib
+import requests
 
 
-_OPENCLIP_S3_BUCKET = 'https://clip-as-service.s3.us-east-2.amazonaws.com/models/torch'
+_OPENCLIP_HUGGINGFACE_BUCKET = 'https://huggingface.co/jinaai/clip-models/'
 _OPENCLIP_MODELS = {
     'RN50::openai': ('RN50.pt', '9140964eaaf9f68c95aa8df6ca13777c'),
     'RN50::yfcc15m': ('RN50-yfcc15m.pt', 'e9c564f91ae7dc754d9043fdcd2a9f22'),
@@ -132,7 +133,7 @@
 def md5file(filename: str):
     hash_md5 = hashlib.md5()
     with open(filename, 'rb') as f:
-        for chunk in iter(lambda: f.read(4096), b""):
+        for chunk in iter(lambda: f.read(4096), b''):
             hash_md5.update(chunk)
 
     return hash_md5.hexdigest()
@@ -143,18 +144,32 @@ def get_model_url_md5(name: str):
     if len(model_pretrained) == 0:  # not on s3
         return None, None
     else:
-        return (_OPENCLIP_S3_BUCKET + '/' + model_pretrained[0], model_pretrained[1])
+        hf_download_url = (
+            _OPENCLIP_HUGGINGFACE_BUCKET
+            + 'resolve/main/'
+            + model_pretrained[0]
+            + '?download=true'
+        )
+        try:
+            response = requests.head(hf_download_url, timeout=10)
+            if response.status_code in [200, 302]:
+                return (hf_download_url, model_pretrained[1])
+        except Exception:
+            raise ValueError('Invalid model url.')
 
 
 def download_model(
     url: str,
-    target_folder: str = os.path.expanduser("~/.cache/clip"),
+    target_folder: str = os.path.expanduser('~/.cache/clip'),
     md5sum: str = None,
     with_resume: bool = True,
     max_attempts: int = 3,
 ) -> str:
     os.makedirs(target_folder, exist_ok=True)
     filename = os.path.basename(url)
+    filename = filename.split('?')[0]
+    if filename.split('.')[-1] == 'onnx':
+        filename = filename.split('-')[-1]
 
     download_target = os.path.join(target_folder, filename)
 
@@ -175,14 +190,14 @@ def download_model(
     )
 
     progress = Progress(
-        " \n",  # divide this bar from Flow's bar
-        TextColumn("[bold blue]{task.fields[filename]}", justify="right"),
-        "[progress.percentage]{task.percentage:>3.1f}%",
-        "•",
+        ' \n',  # divide this bar from Flow's bar
+        TextColumn('[bold blue]{task.fields[filename]}', justify='right'),
+        '[progress.percentage]{task.percentage:>3.1f}%',
+        '•',
         DownloadColumn(),
-        "•",
+        '•',
         TransferSpeedColumn(),
-        "•",
+        '•',
         TimeRemainingColumn(),
     )