Merge branch 'abidlabs/external' of https://github.com/gradio-app/gradio into abidlabs/external

gradio.egg-info/SOURCES.txt

@@ -5,6 +5,7 @@ gradio/__init__.py
 gradio/component.py
 gradio/embeddings.py
 gradio/encryptor.py
+gradio/external.py
 gradio/inputs.py
 gradio/interface.py
 gradio/interpretation.py
@@ -14,6 +15,7 @@ gradio/outputs.py
 gradio/processing_utils.py
 gradio/strings.py
 gradio/test_data.py
+gradio/transforms.py
 gradio/tunneling.py
 gradio/utils.py
 gradio.egg-info/PKG-INFO

gradio/external.py

@@ -128,31 +128,50 @@ def get_huggingface_interface(model_name, api_key, alias):
     return interface_info
 
 def get_gradio_interface(model_name, api_key, alias):
-    api_url = "http://4553.gradiohub.com/api/predict/"  #TODO(dawood): fetch based on model name
-    pipeline = {  #TODO(dawood): load from the config file
-        'inputs': inputs.Textbox(label="Input"),
-        'outputs': outputs.Textbox(label="Question"),
-        'preprocess': lambda x: {"data": [x]},
-        'postprocess': lambda r: r["data"][0],
-        'examples': [['Hi, how are you?']]
+    model_info = requests.get("https://gradio.app/get_config/{}".format(model_name)).json()
+    config_info = json.loads(model_info["config"])
+    api_url = "{}/api/predict/".format(model_info["url"])
+
+    headers = {
+        'authority': model_info["url"],
+        'sec-ch-ua': '" Not A;Brand";v="99", "Chromium";v="90", "Microsoft Edge";v="90"',
+        'accept': 'application/json, text/javascript, */*; q=0.01',
+        'sec-ch-ua-mobile': '?1',
+        'user-agent': 'Mozilla/5.0 (Linux; Android 6.0; Nexus 5 Build/MRA58N) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/90.0.4430.93 Mobile Safari/537.36 Edg/90.0.818.56',
+        'content-type': 'application/json; charset=UTF-8',
+        'origin': 'https://gradio.app',
+        'sec-fetch-site': 'cross-site',
+        'sec-fetch-mode': 'cors',
+        'sec-fetch-dest': 'empty',
+        'referer': 'https://gradio.app/',
+        'accept-language': 'en-US,en;q=0.9',
     }
 
     def query_gradio_api(*input):
         payload = pipeline['preprocess'](*input)
         data = json.dumps(payload)
-        response = requests.request("POST", api_url, data=data)
+        response = requests.post(api_url, headers=headers, data=data)
         result = json.loads(response.content.decode("utf-8"))
         output = pipeline['postprocess'](result)
         return output
 
-    query_gradio_api.__name__ = model_name
+    if alias is None:
+        query_gradio_api.__name__ = model_name
+    else:
+        query_gradio_api.__name__ = alias
+
+    pipeline = {
+        'inputs': [inp[0] for inp in config_info["input_interfaces"]],
+        'outputs': [out[0] for out in config_info["output_interfaces"]],
+        'preprocess': lambda x: {"data": [x]},
+        'postprocess': lambda r: r["data"][0],
+    }
 
     interface_info = {
         'fn': query_gradio_api, 
         'inputs': pipeline['inputs'],
         'outputs': pipeline['outputs'],
         'title': model_name,
-        # 'examples': pipeline['examples'],
     }
 
     return interface_info

gradio/interface.py

@@ -287,6 +287,7 @@ class Interface:
 
             if len(self.output_interfaces) == len(self.predict):
                 prediction = [prediction]
+            
             durations.append(duration)
             predictions.extend(prediction)
         

gradio/transforms.py

@@ -3,34 +3,41 @@ Ways to transform interfaces to produce new interfaces
 """
 from gradio.interface import Interface
 
-def parallel(*interfaces, **options):
-    fns = []
-    outputs = []
-    
-    for io in interfaces:
-        fns.extend(io.predict)
-        outputs.extend(io.output_interfaces)
-    
-    return Interface(fn=fns, inputs=interfaces[0].input_interfaces, outputs=outputs, 
-                     repeat_outputs_per_model=False, **options) 
-
-
-def series(*interfaces, **options):
-    fns = [io.predict for io in interfaces]
-    
-    def connected_fn(data):  # actually not used.
-        for fn in fns:
-            data = fn(data)
-        return data
-    
-    connected_fn.__name__ = " => ".join([f[0].__name__ for f in fns])
-
-    def connected_process_fn(data):  # we have to include the pre/postprocessing of every interface
+class Parallel(Interface):
+    def __init__(self, *interfaces, **options):
+        fns = []
+        outputs = []
+        
         for io in interfaces:
-            data = io.process(data)
-        return data
+            fns.extend(io.predict)
+            outputs.extend(io.output_interfaces)
+    
+        super().__init__(fn=fns, inputs=interfaces[0].input_interfaces, outputs=outputs, 
+                         repeat_outputs_per_model=False, **options) 
 
-    io = Interface(connected_fn, interfaces[0].input_interfaces, interfaces[-1].output_interfaces, **options)
-    io.process = connected_process_fn
-    return io    
+
+class Series(Interface):
+    def __init__(self, *interfaces, **options):
+        fns = [io.predict for io in interfaces]
+    
+        def connected_fn(data):  # Run each function with the appropriate preprocessing and postprocessing 
+            data = [data] # put it in a list before it gets unraveled
+            for idx, io in enumerate(interfaces):
+                # skip preprocessing for first interface since the compound interface will include it
+                if idx > 0:
+                    data = [input_interface.preprocess(data[i]) for i, input_interface in enumerate(io.input_interfaces)]
+                # run all of predictions sequentially
+                predictions = []
+                for predict_fn in io.predict:
+                    prediction = predict_fn(*data)
+                    predictions.append(prediction)
+                data = predictions
+                # skip postprocessing for final interface since the compound interface will include it
+                if idx < len(interfaces) - 1:
+                    data = [output_interface.postprocess(data[i]) for i, output_interface in enumerate(io.output_interfaces)]
+            return data[0]
+    
+        connected_fn.__name__ = " => ".join([f[0].__name__ for f in fns])
+
+        super().__init__(connected_fn, interfaces[0].input_interfaces, interfaces[-1].output_interfaces, **options)