python/nfrom transformers import AutoModel, AutoTokenizer/nimport gradio as gr/nimport mdtex2html/nfrom utils import load_model_on_gpus/n/ntokenizer = AutoTokenizer.from_pretrained('D://work//ChatGLM2-6B//THUDM//chatglm2-6b', trust_remote_code=True)/nmodel = AutoModel.from_pretrained('D://work//ChatGLM2-6B//THUDM//chatglm2-6b', trust_remote_code=True).cuda()/n# 多显卡支持，使用下面两行代替上面一行，将num_gpus改为你实际的显卡数量/n# from utils import load_model_on_gpus/n# model = load_model_on_gpus('THUDM/chatglm2-6b', num_gpus=2)/nmodel = model.eval()/n/n'Override Chatbot.postprocess'/n/ndef postprocess(self, y):/n if y is None:/n return []/n for i, (message, response) in enumerate(y):/n y[i] = (/n None if message is None else mdtex2html.convert((message)),/n None if response is None else mdtex2html.convert(response),/n )/n return y/n/ngr.Chatbot.postprocess = postprocess/n/n/ndef parse_text(text):/n 'copy from https://github.com/GaiZhenbiao/ChuanhuChatGPT/'/n lines = text.split('/n')/n lines = [line for line in lines if line != '']/n count = 0/n for i, line in enumerate(lines):/n if '' in line:/n count += 1/n items = line.split('')/n if count % 2 == 1:/n lines[i] = f'<pre><code class='language-{items[-1]}'>' /n else:/n lines[i] = f'<br></code></pre>'/n else:/n if i > 0:/n if count % 2 == 1:/n line = line.replace('', '/`')/n line = line.replace('<', '<')/n line = line.replace('>', '>')/n line = line.replace(' ', ' ')/n line = line.replace('*', '*')/n line = line.replace('_', '_')/n line = line.replace('-', '-')/n line = line.replace('.', '.')/n line = line.replace('!', '!')/n line = line.replace('(', '(')/n line = line.replace(')', ')')/n line = line.replace('$', '$')/n lines[i] = '
'+line/n text = ''.join(lines)/n return text/n/n/ndef predict(input, chatbot, max_length, top_p, temperature, history, past_key_values):/n chatbot.append((parse_text(input), ''))/n for response, history, past_key_values in model.stream_chat(tokenizer, input, history, past_key_values=past_key_values,/n return_past_key_values=True,/n max_length=max_length, top_p=top_p,/n temperature=temperature):/n chatbot[-1] = (parse_text(input), parse_text(response))/n/n yield chatbot, history, past_key_values/n/n/ndef reset_user_input():/n return gr.update(value='')/n/n/ndef reset_state():/n return [], [], None/n/nwith gr.Blocks() as demo:/n gr.HTML('''

ChatGLM2-6B

''')/n/n chatbot = gr.Chatbot()/n with gr.Row():/n with gr.Column(scale=4):/n with gr.Column(scale=12):/n user_input = gr.Textbox(show_label=False, placeholder='Input...', lines=10).style(/n container=False)/n with gr.Column(min_width=32, scale=1):/n submitBtn = gr.Button('Submit', variant='primary')/n with gr.Column(scale=1):/n emptyBtn = gr.Button('Clear History')/n max_length = gr.Slider(0, 32768, value=8192, step=1.0, label='Maximum length', interactive=True)/n top_p = gr.Slider(0, 1, value=0.8, step=0.01, label='Top P', interactive=True)/n temperature = gr.Slider(0, 1, value=0.95, step=0.01, label='Temperature', interactive=True)/n/n history = gr.State([])/n past_key_values = gr.State(None)/n/n submitBtn.click(predict, [user_input, chatbot, max_length, top_p, temperature, history, past_key_values],/n [chatbot, history, past_key_values], show_progress=True)/n submitBtn.click(reset_user_input, [], [user_input])/n/n emptyBtn.click(reset_state, outputs=[chatbot, history, past_key_values], show_progress=True)/n/ndemo.queue().launch(share=False, inbrowser=True)/n```