add app

app.py

@@ -0,0 +1,109 @@
+# import streamlit as st
+# from streamlit_chat import message as st_message
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+MAX_HISTORY = 7
+MODEL_PATH = 'llongpre/DialoGPT-small-miles'
+
+def get_models():
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+    model = AutoModelForCausalLM.from_pretrained(MODEL_PATH)
+    return tokenizer, model
+
+# if "history" not in st.session_state:
+#     st.session_state.history = []
+#
+# if "history_ids" not in st.session_state:
+#     st.session_state.history_ids = []
+#
+# st.title("Chat with me")
+
+# def generate_answer():
+#     tokenizer, model = get_models()
+#     user_message = st.session_state.input_text
+#     new_user_input_ids = tokenizer.encode(st.session_state.input_text + tokenizer.eos_token, return_tensors='pt')
+#     st.session_state.history_ids.append(new_user_input_ids)
+#     if len(st.session_state.history_ids) > MAX_HISTORY:
+#         st.session_state.history_ids = st.session_state.history_ids[-MAX_HISTORY:]
+#     bot_input_ids = torch.cat(st.session_state.history_ids, dim=-1)
+#     chat_history_ids = model.generate(
+#         bot_input_ids,
+#         pad_token_id=tokenizer.pad_token_id,
+#         max_length=1000,
+#         do_sample=True,
+# #         top_k=150, # sample from the top k words sorted descending by probability
+#         top_p=0.7, # choose smallest possible words whose cumulative probability exceeds p
+#         temperature = 0.95, # 0 greedy, inf is random
+#         no_repeat_ngram_size=3,
+#     )
+#     response = chat_history_ids[:, bot_input_ids.shape[-1]:]
+#     st.session_state.history_ids.append(response)
+#     output = tokenizer.decode(response[0], skip_special_tokens=True)
+#
+#     st.session_state.history.append({"message": user_message, "is_user": True})
+#     st.session_state.history.append({"message": output, "is_user": False})
+
+# st.text_input("Your text message", key="input_text", on_change=generate_answer, placeholder='')
+
+# for chat in st.session_state.history:
+#     st_message(**chat)  # unpacking
+
+
+import gradio as gr
+
+tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+model = AutoModelForCausalLM.from_pretrained(MODEL_PATH)
+
+def predict(input, history=[]):
+    # tokenize the new input sentence
+    new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
+
+    # append the new user input tokens to the chat history
+    bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
+
+    # generate a response
+    history = model.generate(
+                            bot_input_ids,
+                            max_length=1000,
+                            pad_token_id=tokenizer.eos_token_id,
+                            no_repeat_ngram_size=3,
+                            top_p = 0.92,
+                            top_k = 50
+                            ).tolist()
+
+    # convert the tokens to text, and then split the responses into lines
+    response = tokenizer.decode(history[0]).split("<|endoftext|>")
+    response = [(response[i], response[i+1]) for i in range(0, len(response)-1, 2)]  # convert to tuples of list
+
+    return response, history
+
+def generate_answer(input, history=[]):
+    new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
+    # st.session_state.history_ids.append(new_user_input_ids)
+    history = history.append(new_user_input_ids)
+    if len(history) > MAX_HISTORY:
+        history = history[-MAX_HISTORY:]
+    bot_input_ids = torch.cat(history, dim=-1)
+    chat_history_ids = model.generate(
+        bot_input_ids,
+        pad_token_id=tokenizer.pad_token_id,
+        max_length=1000,
+        do_sample=True,
+#         top_k=150, # sample from the top k words sorted descending by probability
+        top_p=0.7, # choose smallest possible words whose cumulative probability exceeds p
+        temperature = 0.95, # 0 greedy, inf is random
+        no_repeat_ngram_size=3,
+    )
+    response = chat_history_ids[:, bot_input_ids.shape[-1]:]
+    history.append(response)
+    output = tokenizer.decode(response[0], skip_special_tokens=True)
+
+    return output, history
+
+
+gr.Interface(fn=generate_answer,
+    title="DialoGPT-large",
+    inputs=["text", "state"],
+    outputs=["chatbot", "state"],
+    ).launch()

requirements.txt

@@ -0,0 +1,2 @@
+transformers
+torch