Update app.py

app.py

@@ -1,33 +1,3 @@
-
-Hugging Face's logo Hugging Face
-
-Models
-Datasets
-Spaces
-Docs
-Pricing
-
-Spaces:
-bleysg
-/
-WizardCoder-Python-34b-v1.0
-private
-App
-Files
-Community
-Settings
-WizardCoder-Python-34b-v1.0
-/ app.py
-bleysg's picture
-bleysg
-Update app.py
-7a84795
-about 1 hour ago
-raw
-history
-blame
-No virus
-5.52 kB
 import os
 import re
 import logging
@@ -35,13 +5,20 @@ import gradio as gr
 import openai
 
 print(os.environ)
-openai.api_base = os.environ.get("OPENAI_API_BASE")
+openai.api_base1 = os.environ.get("OPENAI_API_BASE")
+openai.api_base2 = os.environ.get("OPENAI_API_BASE2")
 openai.api_key = os.environ.get("OPENAI_API_KEY")
 
 BASE_SYSTEM_MESSAGE = """"""
 
-def make_prediction(prompt, max_tokens=None, temperature=None, top_p=None, top_k=None, repetition_penalty=None):
-    completion = openai.Completion.create(model="/workspace/text-generation-webui/models/wizardcoder-python-34b-v1.0.Q5_K_M.gguf", prompt=prompt, max_tokens=max_tokens, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty, stream=True, stop=["</s>", "<|im_end|>"])
+
+def make_prediction1(prompt, max_tokens=None, temperature=None, top_p=None, top_k=None, repetition_penalty=None):
+    completion = openai.Completion.create(api_base=openai.api_base1, model="wizardcoder-python-34b-v1.0.Q5_K_M.gguf", prompt=prompt, max_tokens=max_tokens, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty, stream=True, stop=["</s>", "<|im_end|>"])
+    for chunk in completion:
+        yield chunk["choices"][0]["text"]
+
+def make_prediction2(prompt, max_tokens=None, temperature=None, top_p=None, top_k=None, repetition_penalty=None):
+    completion = openai.Completion.create(api_base=openai.api_base2, model="wizardcoder-python-34b-v1.0.Q5_K_M.gguf", prompt=prompt, max_tokens=max_tokens, temperature=temperature, top_p=top_p, top_k=top_k, repetition_penalty=repetition_penalty, stream=True, stop=["</s>", "<|im_end|>"])
     for chunk in completion:
         yield chunk["choices"][0]["text"]
 
@@ -59,7 +36,7 @@ def user(message, history):
     return "", history
 
 
-def chat(history, system_message, max_tokens, temperature, top_p, top_k, repetition_penalty):
+def chat1(history, system_message, max_tokens, temperature, top_p, top_k, repetition_penalty):
     history = history or []
 
     messages = BASE_SYSTEM_MESSAGE + system_message.strip() + "\n" + \
@@ -70,7 +47,38 @@ def chat(history, system_message, max_tokens, temperature, top_p, top_k, repetit
     # remove last space from assistant, some models output a ZWSP if you leave a space
     messages = messages.rstrip()
 
-    prediction = make_prediction(
+    prediction = make_prediction1(
+        messages,
+        max_tokens=max_tokens,
+        temperature=temperature,
+        top_p=top_p,
+        top_k=top_k,
+        repetition_penalty=repetition_penalty,
+    )
+    for tokens in prediction:
+        tokens = re.findall(r'(.*?)(\s|$)', tokens)
+        for subtoken in tokens:
+            subtoken = "".join(subtoken)
+            # Remove "Response\n" if it's at the beginning of the assistant's output
+            if subtoken.startswith("Response"):
+                subtoken = subtoken[len("Response"):]
+            answer = subtoken
+            history[-1][1] += answer
+            # stream the response
+            yield history, history, ""
+
+def chat2(history, system_message, max_tokens, temperature, top_p, top_k, repetition_penalty):
+    history = history or []
+
+    messages = BASE_SYSTEM_MESSAGE + system_message.strip() + "\n" + \
+               "\n".join(["\n".join(["###Instruction\n"+item[0]+"\n\n", "###Response\n"+item[1]+"\n\n"])
+                          for item in history])
+    # strip the last `<|end_of_turn|>` from the messages
+    #messages = messages.rstrip("<|end_of_turn|>")
+    # remove last space from assistant, some models output a ZWSP if you leave a space
+    messages = messages.rstrip()
+
+    prediction = make_prediction2(
         messages,
         max_tokens=max_tokens,
         temperature=temperature,
@@ -141,15 +149,20 @@ with gr.Blocks(css=CSS) as demo:
 
     chat_history_state = gr.State()
     clear.click(clear_chat, inputs=[chat_history_state, message], outputs=[chat_history_state, message], queue=False)
-    clear.click(lambda: None, None, chatbot_gguf, queue=False)
-    clear.click(lambda: None, None, chatbot_gguf2, queue=False)
+    clear.click(lambda: None, None, chatbot1, queue=False)
+    clear.click(lambda: None, None, chatbot2, queue=False)
 
-    submit_click_event = submit.click(
+    submit_click_event1 = submit.click(
+        fn=user, inputs=[message, chat_history_state], outputs=[message, chat_history_state], queue=True
+    ).then(
+        fn=chat1, inputs=[chat_history_state, system_msg, max_tokens, temperature, top_p, top_k, repetition_penalty], outputs=[chatbot1, chat_history_state, message], queue=True
+    )
+    submit_click_event2 = submit.click(
         fn=user, inputs=[message, chat_history_state], outputs=[message, chat_history_state], queue=True
     ).then(
-        fn=chat, inputs=[chat_history_state, system_msg, max_tokens, temperature, top_p, top_k, repetition_penalty], outputs=[chatbot_gguf, chat_history_state, message], queue=True
+        fn=chat2, inputs=[chat_history_state, system_msg, max_tokens, temperature, top_p, top_k, repetition_penalty], outputs=[chatbot1, chat_history_state, message], queue=True
     )
-    stop.click(fn=None, inputs=None, outputs=None, cancels=[submit_click_event], queue=False)
+    stop.click(fn=None, inputs=None, outputs=None, cancels=[submit_click_event1, submit_click_event2], queue=False)
 
 demo.queue(max_size=48, concurrency_count=8).launch(debug=True, server_name="0.0.0.0", server_port=7860)