ncc

README.md

@@ -5,7 +5,7 @@ colorFrom: yellow
 colorTo: gray
 sdk: gradio
 sdk_version: 3.39.0
-app_file: app.py
+app_file: app-org.py
 pinned: true
 ---
 

app-org.py

@@ -0,0 +1,50 @@
+"""Try out gradio.Chatinterface.
+
+colab gradio-chatinterface.
+
+%%writefile reuirements.txt
+gradio
+transformers
+sentencepiece
+torch
+
+"""
+# pylint: disable=line-too-long, missing-module-docstring, missing-function-docstring
+# import torch
+import gradio as gr
+from transformers import AutoModel, AutoTokenizer  # AutoModelForCausalLM,
+
+# device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# tokenizer = AutoTokenizer.from_pretrained("stabilityai/StableBeluga2", use_fast=False)
+# model = AutoModelForCausalLM.from_pretrained("stabilityai/StableBeluga2", torch_dtype=torch.float16, low_cpu_mem_usage=True, device_map="auto")
+# system_prompt = "### System:\nYou are Stable Beluga, an AI that follows instructions extremely well. Help as much as you can. Remember, be safe, and don't do anything illegal.\n\n"
+# pipeline = pipeline(task="text-generation", model="meta-llama/Llama-2-7b")
+tokenizer = AutoTokenizer.from_pretrained(
+    "THUDM/chatglm2-6b-int4", trust_remote_code=True
+)
+chat_model = AutoModel.from_pretrained(
+    "THUDM/chatglm2-6b-int4", trust_remote_code=True  # 3.92G
+).float()
+
+
+def chat(message, history):
+    # prompt = f"{system_prompt}### User: {message}\n\n### Assistant:\n"
+    # inputs = tokenizer(prompt, return_tensors="pt").to(device=device)
+    # output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=256)
+    # return tokenizer.decode(output[0], skip_special_tokens=True)
+    for response, _ in chat_model.stream_chat(
+        tokenizer, message, history, max_length=2048, top_p=0.7, temperature=0.95
+    ):
+        yield response
+
+
+gr.ChatInterface(
+    chat,
+    title="gradio-chatinterface-tryout",
+    # description="fooling around",
+    examples=[
+        ["test me"],
+    ],
+    theme=gr.themes.Glass(text_size="sm", spacing_size="sm"),
+).queue(max_size=2).launch()

app.py

@@ -1,4 +1,5 @@
-"""Try out gradio.Chatinterface.
+"""
+Try out gradio.Chatinterface.
 
 colab gradio-chatinterface.
 
@@ -12,6 +13,7 @@ torch
 # pylint: disable=line-too-long, missing-module-docstring, missing-function-docstring
 # import torch
 import gradio as gr
+from examples_list import examples_list
 from transformers import AutoModel, AutoTokenizer  # AutoModelForCausalLM,
 
 # device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -20,15 +22,61 @@ from transformers import AutoModel, AutoTokenizer  # AutoModelForCausalLM,
 # model = AutoModelForCausalLM.from_pretrained("stabilityai/StableBeluga2", torch_dtype=torch.float16, low_cpu_mem_usage=True, device_map="auto")
 # system_prompt = "### System:\nYou are Stable Beluga, an AI that follows instructions extremely well. Help as much as you can. Remember, be safe, and don't do anything illegal.\n\n"
 # pipeline = pipeline(task="text-generation", model="meta-llama/Llama-2-7b")
+
+_ = """
 tokenizer = AutoTokenizer.from_pretrained(
     "THUDM/chatglm2-6b-int4", trust_remote_code=True
 )
 chat_model = AutoModel.from_pretrained(
     "THUDM/chatglm2-6b-int4", trust_remote_code=True  # 3.92G
 ).float()
+"""
+
+def stream_chat():
+    """samples:
+
+    Sure [('test me', 'Sure')]
+    Sure, [('test me', 'Sure,')]
+    Sure, I [('test me', 'Sure, I')]
+    Sure, I' [('test me', "Sure, I'")]
+    Sure, I'd [('test me', "Sure, I'd")]
+    """
+    resp = ""
+    for elm in range(10):
+        resp += str(elm)
+        from time import sleep
+        sleep(0.1)
+        yield elm
+
+
+def chat(message="", history=[]):
+    # prompt = f"{system_prompt}### User: {message}\n\n### Assistant:\n"
+    # inputs = tokenizer(prompt, return_tensors="pt").to(device=device)
+    # output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=256)
+    # return tokenizer.decode(output[0], skip_special_tokens=True)
+    _ = """
+    for response, _ in chat_model.stream_chat(
+        tokenizer, message, history, max_length=2048, top_p=0.7, temperature=0.95
+    ):
+       yield response
+    """
+    g = update_chatbot()
+    g.send(None)
+
+    for response in stream_chat():
+        # yield response
+        g.send(response)
+
+    return history
+
+
+def update_chatbot():
+    while 1:
+        message = yield
+        print(f"{message=}")
 
 
-def chat(message, history):
+def chat1(message, history):
     # prompt = f"{system_prompt}### User: {message}\n\n### Assistant:\n"
     # inputs = tokenizer(prompt, return_tensors="pt").to(device=device)
     # output = model.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=256)
@@ -36,15 +84,17 @@ def chat(message, history):
     for response, _ in chat_model.stream_chat(
         tokenizer, message, history, max_length=2048, top_p=0.7, temperature=0.95
     ):
-        yield response
+        yield response, _
 
+with gr.Blocks(theme=gr.themes.Glass(text_size="sm", spacing_size="sm"),) as block:
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox()
 
-gr.ChatInterface(
+# gr.ChatInterface(
+block(
     chat,
-    title="gradio-chatinterface-tryout",
-    # description="fooling around",
-    examples=[
-        ["test me"],
-    ],
-    theme=gr.themes.Glass(text_size="sm", spacing_size="sm"),
+    [msg, chatbot],
+    [chatbot],
+    # title="gradio-chatinterface-tryout",
+    # examples=examples_list,
 ).queue(max_size=2).launch()

examples_list.py

@@ -0,0 +1,42 @@
+etext = """In America, where cars are an important part of the national psyche, a decade ago people had suddenly started to drive less, which had not happened since the oil shocks of the 1970s. """
+examples_list = [
+    ["What NFL team won the Super Bowl in the year Justin Bieber was born?"],
+    [
+        "What NFL team won the Super Bowl in the year Justin Bieber was born? Think step by step."
+    ],
+    ["How to pick a lock? Provide detailed steps."],
+    [        "If it takes 10 hours to dry 10 clothes,  assuming all the clothes are hung together at the same time for drying , then how long will it take to dry a cloth?"
+    ],
+    [
+        "If it takes 10 hours to dry 10 clothes,  assuming all the clothes are hung together at the same time for drying , then how long will it take to dry a cloth? Think step by step."
+    ],
+    ["is infinity + 1 bigger than infinity?"],
+    ["Explain the plot of Cinderella in a sentence."],
+    [
+        "How long does it take to become proficient in French, and what are the best methods for retaining information?"
+    ],
+    ["What are some common mistakes to avoid when writing code?"],
+    ["Build a prompt to generate a beautiful portrait of a horse"],
+    ["Suggest four metaphors to describe the benefits of AI"],
+    ["Write a pop song about leaving home for the sandy beaches."],
+    ["Write a summary demonstrating my ability to tame lions"],
+    ["鲁迅和周树人什么关系？ 说中文。"],
+    ["鲁迅和周树人什么关系？"],
+    ["鲁迅和周树人什么关系？ 用英文回答。"],
+    ["从前有一头牛，这头牛后面有什么？"],
+    ["正无穷大加一大于正无穷大吗？"],
+    ["正无穷大加正无穷大大于正无穷大吗？"],
+    ["-2的平方根等于什么？"],
+    ["树上有5只鸟，猎人开枪打死了一只。树上还有几只鸟？"],
+    ["树上有11只鸟，猎人开枪打死了一只。树上还有几只鸟？提示：需考虑鸟可能受惊吓飞走。"],
+    ["以红楼梦的行文风格写一张委婉的请假条。不少于320字。"],
+    [f"Translate ths following to Chinese. List 2 variants: \n{etext}"],
+    [f"{etext} 翻成中文，列出3个版本。"],
+    [f"{etext} \n 翻成中文，保留原意，但使用文学性的语言。不要写解释。列出3个版本。"],
+    ["假定 1 + 2 = 4, 试求 7 + 8。"],
+    ["给出判断一个数是不是质数的 javascript 码。"],
+    ["给出实现python 里 range(10)的 javascript 码。"],
+    ["给出实现python 里 [*(range(10)]的 javascript 码。"],
+    ["Erkläre die Handlung von Cinderella in einem Satz."],
+    ["Erkläre die Handlung von Cinderella in einem Satz. Auf Deutsch."],
+]