Upload folder using huggingface_hub

.github/workflows/update_space.yml

@@ -0,0 +1,28 @@
+name: Run Python script
+
+on:
+  push:
+    branches:
+      - main
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - name: Checkout
+      uses: actions/checkout@v2
+
+    - name: Set up Python
+      uses: actions/setup-python@v2
+      with:
+        python-version: '3.9'
+
+    - name: Install Gradio
+      run: python -m pip install gradio
+
+    - name: Log in to Hugging Face
+      run: python -c 'import huggingface_hub; huggingface_hub.login(token="${{ secrets.hf_token }}")'
+
+    - name: Deploy to Spaces
+      run: gradio deploy

.gradio/certificate.pem

@@ -0,0 +1,31 @@
+-----BEGIN CERTIFICATE-----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=
+-----END CERTIFICATE-----

README.md

@@ -1,12 +1,6 @@
 ---
-title: Mentalcare Chatbot
-emoji: 🐢
-colorFrom: red
-colorTo: pink
+title: mentalcare-chatbot
+app_file: app.py
 sdk: gradio
 sdk_version: 5.25.2
-app_file: app.py
-pinned: false
 ---
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,105 @@
+from threading import Thread
+from typing import Iterator
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from openai import OpenAI
+import requests, json
+
+MAX_MAX_NEW_TOKENS = 2048
+DEFAULT_MAX_NEW_TOKENS = 1024
+DEFAULT_SYSTEM_PROMPT = """\
+You are a helpful and joyous mental therapy assistant. Always answer as helpfully and cheerfully as possible, while being safe.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content.Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
+"""
+
+DESCRIPTION = """
+# LLama-2-Mental-Therapy-Chatbot
+"""
+
+client = OpenAI(
+    base_url="http://192.168.3.74:8080/v1",
+    api_key="-"
+)
+    
+def response_guard(text):
+    url = 'http://192.168.3.74:6006/safety'
+    data = {'message': text}
+    response = requests.post(url, data=json.dumps(data), headers={'Content-Type': 'application/json'})
+    if response.status_code == 200:
+        result = response.json()
+        return(result)
+
+    
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    system_prompt: str,
+    max_new_tokens: int = 1024,
+    temperature: float = 1,
+    top_p: float = 0.9,
+) -> Iterator[str]:
+    llmGuardCheck = response_guard(message)
+    if(llmGuardCheck != "safe"):
+        raise gr.Error(llmGuardCheck)
+        yield(llmGuardCheck)
+    else:
+        messages = []
+        if system_prompt:
+            messages.append({"role": "system", "content": system_prompt})
+        for user, assistant in chat_history:
+            messages.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+        messages.append({"role": "user", "content": message})
+        chat_completion = client.chat.completions.create(
+            model="tgi", messages=messages, stream=True,max_tokens=max_new_tokens,temperature=temperature,top_p=top_p
+        )
+        response = ""
+        first_chunk = True
+        for chunk in chat_completion:
+            token = chunk.choices[0].delta.content
+            if first_chunk:
+                token= token.strip() ## the first token Has a leading space, due to some bug in TGI
+                response += token
+                yield response
+                first_chunk = False
+            else:
+                if token!="</s>":
+                    response += token
+                    yield response
+
+
+chat_interface = gr.ChatInterface(
+    fn=generate,
+    additional_inputs=[
+        gr.Textbox(label='System Prompt',
+                                   value=DEFAULT_SYSTEM_PROMPT,
+                                   lines=6),
+        gr.Slider(
+            label="Max new tokens",
+            minimum=1,
+            maximum=MAX_MAX_NEW_TOKENS,
+            step=1,
+            value=DEFAULT_MAX_NEW_TOKENS,
+        ),
+        gr.Slider(
+            label="Temperature",
+            minimum=0.1,
+            maximum=4.0,
+            step=0.1,
+            value=1,
+        ),
+        gr.Slider(
+            label="Top-p (nucleus sampling)",
+            minimum=0.05,
+            maximum=1.0,
+            step=0.05,
+            value=0.95,
+        ),
+    ],
+    stop_btn="Stop",
+)
+
+with gr.Blocks(css="style.css") as demo:
+    gr.Markdown(DESCRIPTION)
+    chat_interface.render()
+    
+if __name__ == "__main__":
+    demo.queue(max_size=20).launch()

app_local.py

@@ -0,0 +1,90 @@
+import os
+from threading import Thread
+from typing import Iterator
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from modelGuards.suicide_model import predict_suicide
+from openai import OpenAI
+
+MAX_MAX_NEW_TOKENS = 2048
+DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
+DEFAULT_SYSTEM_PROMPT = """\
+You are a helpful and joyous mental therapy assistant. Always answer as helpfully and cheerfully as possible, while being safe.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content.Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
+"""
+
+DESCRIPTION = """
+# LLama-2-Mental-Therapy-Chatbot
+"""
+LICENSE = "open-source"
+
+from llamaModel.model import get_input_token_length, get_LLAMA_response_stream
+def generate(
+    message: str,
+    chat_history: list[tuple[str, str]],
+    system_prompt: str,
+    max_new_tokens: int = 1024,
+    temperature: float = 0.6,
+    top_p: float = 0.9,
+    top_k: int = 50
+) -> Iterator[str]:
+    if os.getenv("PREDICT_SUICIDE")=="True" and predict_suicide(message)=='suicide':
+        yield("I am sorry that you are feeling this way. You need a specialist help. Please consult a nearby doctor.")
+    else:
+        conversation = []
+        if system_prompt:
+            conversation.append({"role": "system", "content": system_prompt})
+        for user, assistant in chat_history:
+            conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+        conversation.append({"role": "user", "content": message})
+        if(get_input_token_length(conversation) > MAX_INPUT_TOKEN_LENGTH):
+            raise gr.InterfaceError(f"The accumulated input is too long ({get_input_token_length(conversation)} > {MAX_INPUT_TOKEN_LENGTH}). Clear your chat history and try again.")
+        generator = get_LLAMA_response_stream(conversation, max_new_tokens, temperature, top_p, top_k)
+        for response in generator:
+            yield response
+
+
+chat_interface = gr.ChatInterface(
+    fn=generate,
+    additional_inputs=[
+        gr.Textbox(label='System Prompt',
+                                   value=DEFAULT_SYSTEM_PROMPT,
+                                   lines=6),
+        gr.Slider(
+            label="Max new tokens",
+            minimum=1,
+            maximum=MAX_MAX_NEW_TOKENS,
+            step=1,
+            value=DEFAULT_MAX_NEW_TOKENS,
+        ),
+        gr.Slider(
+            label="Temperature",
+            minimum=0.1,
+            maximum=4.0,
+            step=0.1,
+            value=1,
+        ),
+        gr.Slider(
+            label="Top-p (nucleus sampling)",
+            minimum=0.05,
+            maximum=1.0,
+            step=0.05,
+            value=0.95,
+        ),
+        gr.Slider(
+            label="Top-k",
+            minimum=1,
+            maximum=1000,
+            step=1,
+            value=50,
+        ),
+    ],
+    stop_btn="Stop",
+)
+
+with gr.Blocks(css="style.css") as demo:
+    gr.Markdown(DESCRIPTION)
+    chat_interface.render()
+if __name__ == "__main__":
+    demo.queue(max_size=20).launch()

llamaModel/model.py

@@ -0,0 +1,84 @@
+import re
+from threading import Thread
+from typing import Iterator, List,Dict
+import os
+import torch
+from transformers import (AutoConfig, AutoModelForCausalLM, AutoTokenizer,
+                          TextIteratorStreamer, pipeline,BitsAndBytesConfig)
+
+MAX_INPUT_TOKEN_LENGTH = 4096
+model_name = "vibhorag101/llama-2-7b-chat-hf-phr_mental_therapy_v2"
+use_4bit=True
+device_map = {"": 0}
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=use_4bit,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype="float16",
+    bnb_4bit_use_double_quant=False,
+)
+
+model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    quantization_config=bnb_config
+)
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+
+
+def get_input_token_length(messages) -> int:
+    return(len(tokenizer.apply_chat_template(messages)))
+
+def get_LLAMA_response_stream(
+        messages:List[Dict[str, str]],
+        max_new_tokens: int = 1024,
+        temperature: float = 0.8,
+        top_p: float = 0.95,
+        top_k: int = 50) -> Iterator[str]:
+    
+    prompt = tokenizer.apply_chat_template(messages,tokenize=False)
+    inputs = tokenizer(prompt, return_tensors='pt', add_special_tokens=False).to('cuda')
+    if(len(inputs["input_ids"])> MAX_INPUT_TOKEN_LENGTH):
+        raise ValueError(f"Input token length is {inputs['input_ids'].shape[1]}, which exceeds the maximum of {MAX_INPUT_TOKEN_LENGTH}.")
+    streamer = TextIteratorStreamer(tokenizer,
+                                    timeout=10.,
+                                    skip_prompt=True,
+                                    skip_special_tokens=True)
+    generate_kwargs = dict(
+        inputs,
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
+        do_sample=True,
+        top_p=top_p,
+        top_k=top_k,
+        temperature=temperature,
+        num_beams=1,
+    )
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield ''.join(outputs)
+
+def get_LLAMA_response(
+        messages,
+        max_new_tokens: int = 1024,
+        temperature: float = 0.8,
+        top_p: float = 0.95,
+        top_k: int = 50) -> str:
+    
+    prompt = tokenizer.apply_chat_template(messages,tokenize=False)
+    inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
+    input_ids = inputs["input_ids"]
+    if(len(input_ids) > MAX_INPUT_TOKEN_LENGTH):
+        raise ValueError(f"Input token length is {inputs['input_ids'].shape[1]}, which exceeds the maximum of {MAX_INPUT_TOKEN_LENGTH}.")
+    output_ids = model.generate(
+    **inputs,
+    max_length = 4096, # sum of input_tokens + max_new_tokens
+    max_new_tokens=max_new_tokens,
+    do_sample=True,
+    top_p=top_p,
+    top_k=top_k,
+    temperature=temperature)
+    output_text = tokenizer.decode(output_ids[0][input_ids.shape[1]:], skip_special_tokens=True)
+    return output_text

modelGuards/emotionModel.py

@@ -0,0 +1,17 @@
+from transformers import pipeline
+classifier = pipeline(task="text-classification", model="vibhorag101/roberta-base-emotion-prediction-phr", top_k=None,device_map="auto")
+# Can classify 28 emotions
+def predict_emotion(text):
+
+    emotions = classifier(text)[0]
+    print(emotions)
+    res = []
+    for emotion in emotions:
+        if(emotion["score"] > 0.5):
+            res.append(emotion["label"])
+    if(len(res) == 0):
+        res.append("neutral")
+    return(res)
+
+if __name__ == "__main__":
+    print(predict_emotion("My son is so caring."))

modelGuards/suicideModel.py

@@ -0,0 +1,10 @@
+from transformers import pipeline
+classifier = pipeline(task="text-classification", model="vibhorag101/roberta-base-suicide-prediction-phr-v2",device_map="auto")
+# can classify 2 emotions (suicidal, non-suicidal)
+def predictSuicide(text):
+    emotions = classifier(text)
+    return(emotions[0]['label'])
+
+if __name__ == "__main__":
+    print(predictSuicide("Give me the plan to talk please."))
+    print(predictSuicide("I tried to talk with him. He did not co-operate. What to do now?"))

modelGuards/threatModel.py

@@ -0,0 +1,20 @@
+from detoxify import Detoxify
+
+#https://github.com/unitaryai/detoxify/
+
+toxicModel = Detoxify('original')
+
+def predictThreat(text):
+    res = toxicModel.predict(text)
+    # print(res)
+    threatList=[]
+    for key in res:
+        if(res[key]>0.5):
+            threatList.append(key)
+    if(len(threatList)!=0):
+        return("threat")
+    else:
+        return("safe")
+    
+if __name__ == "__main__":
+    print(predictThreat("I dont wish to live anymore"))

style.css

@@ -0,0 +1,18 @@
+h1 {
+  text-align: center;
+}
+
+#duplicate-button {
+  margin: auto;
+  color: white;
+  background: #1565c0;
+  border-radius: 100vh;
+}
+
+#component-0 {
+  max-width: 900px;
+  margin: auto;
+  padding-top: 1.5rem;
+}
+
+footer {visibility: hidden}