mindpadi
/

hybrid_classifier_suite

@@ -2,269 +2,128 @@
 license: mit
 language:
   - en
-base_model:
-  - openai-community/gpt2
-  - distilbert/distilgpt2
-pipeline_tag: text-generation
 tags:
-  - medical
 ---
-# 🧠 Model Card: MindPadi Models
-This model card documents the machine learning models developed for **MindPadi**, a mental health chatbot offering conversational support, emotion detection, intent classification, and therapy-aligned features. The models power tasks such as text generation, sentiment/emotion analysis, and user intent recognition.
-## 📑 Table of Contents
-- [Model Overview](#model-overview)
-- [Models](#models)
-- [Intended Use](#intended-use)
-- [Training Details](#training-details)
-- [Evaluation](#evaluation)
-- [Limitations](#limitations)
-- [Usage Instructions](#usage-instructions)
-- [Ethical Considerations](#ethical-considerations)
-- [Contact](#contact)
-## 🧬 Model Overview
-MindPadi uses a combination of pre-trained and fine-tuned transformer models for:
-- 🗣️ Text generation
-- 🎭 Emotion detection
-- 🎯 Intent classification
-- 📊 Sentence embedding
-Models are stored in the `models/` directory and published to the Hugging Face Hub for scalable inference.
-## 🧠 Models
-### 🔹 `distilgpt2`
-- **Description**: Pre-trained DistilGPT-2 for text generation
-- **Task**: Text Generation
-- **Architecture**: 6-layer transformer (82M params)
-- **Use Case**: Default conversational model
-### 🔹 `fine_tuned_distilgpt2_lora`
-- **Description**: DistilGPT-2 fine-tuned using LoRA for mental health contexts
-- **Training Script**: `training/finetune_distilgpt2_lora.py`
-- **Use Case**: Therapy-specific generation
-### 🔹 `fine_tuned_gpt2`
-- **Description**: GPT-2 fine-tuned for rich and context-aware dialogues
-- **Architecture**: GPT-2 (124M params)
-- **Training Script**: `training/finetune_gpt2_pipeline.py`
-### 🔹 `merged_distilgpt2`
-- **Description**: Optimized DistilGPT-2 with merged fine-tuned weights
-- **Use Case**: Fallback generation model
-### 🔹 `gpt2`
-- **Description**: Raw GPT-2 as a baseline model
-- **Source**: Hugging Face Transformers
-### 🔹 `emotion_classifier`
-- **Task**: Emotion Classification (e.g., joy, sadness, anger)
-- **Training Script**: `training/train_emotion_model.py`
-- **Use Case**: Used in `app/chatbot/emotion.py`
-### 🔹 `emotion_model`
-- **Description**: Variant or backup for emotion analysis
-### 🔹 `intent_classifier`
-- **Task**: User intent detection (e.g., schedule, vent, help)
-- **Training Script**: `training/train_intent_classifier.py`
-- **Use Case**: `app/chatbot/intent_classifier.py`
-### 🔹 `intent_encoder`
-- **Description**: Sentence-BERT used to embed user input
-- **Use Case**: Vector search in `app/utils/embedding_search.py`
-### 🔹 `intent_fallback`
-- **Description**: Fallback model for intent classification errors
-### 🔹 `sentence_transformer`
-- **Architecture**: Sentence-BERT (e.g., all-MiniLM-L6-v2)
-- **Use Case**: Text embedding for similarity queries
-## 🎯 Intended Use
-These models are intended for use in:
-- Conversational therapy interfaces
-- Mental health chatbots
-- Emotion-aware agents
-- Intent-based routing systems
-### 👥 Primary Users:
-- End-users of the MindPadi mental health app
-- Developers integrating AI into mental health tools
-### 🚫 Out-of-Scope:
-- Medical diagnosis
-- Legal/financial decision-making
-- Non-mental health chatbots without validation
-## 🛠 Training Details
-### 🧾 Datasets
-- Location: `training/datasets/`
-- Intents: Stored in `data/processed_intents.json`
-- Processing scripts: `process_conversation_data.py`, `convert_intents_format.py`
-### 💻 Environment
-- Hardware: NVIDIA GPUs (local/cloud)
-- Pretrained models: from Hugging Face
-- Fine-tuned models: custom scripts
-### 🔧 Scripts Used
-| Model | Script |
-|-------|--------|
-| LoRA GPT-2 | `training/finetune_distilgpt2_lora.py` |
-| Fine-tuned GPT-2 | `training/finetune_gpt2_pipeline.py` |
-| Emotion Classifier | `training/train_emotion_model.py` |
-| Intent Classifier | `training/train_intent_classifier.py` |
-## 📊 Evaluation
-### ✅ Metrics
-- **Text Gen**: Perplexity, BLEU
-- **Classification**: Accuracy, F1-score
-### 📈 Results
-- Emotion classifier: High accuracy
-- Fine-tuned GPT models: Better than baseline
-- Evaluation logs: `logs/training.log`, TensorBoard
-## ⚠ Limitations
-- **Bias**: Possible due to training data
-- **Generalization**: May fail on out-of-domain text
-- **Language**: Only English supported
-- **Inference Cost**: Large models require GPU memory
-- **Safety**: Human monitoring is recommended
-## 🚀 Usage Instructions
-### 🔧 Prerequisites
-- Python 3.10+
-- Install:
-  ```bash
-  pip install transformers huggingface_hub requests
-  ```
-### ✍️ Example: Text Generation
 ```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-model = AutoModelForCausalLM.from_pretrained("mindpadi/distilgpt2")
-tokenizer = AutoTokenizer.from_pretrained("mindpadi/distilgpt2")
-inputs = tokenizer("How are you feeling today?", return_tensors="pt")
-outputs = model.generate(**inputs, max_length=50)
-print(tokenizer.decode(outputs[0], skip_special_tokens=True))
 ```
-### 😢 Example: Emotion Classification
-```python
-from transformers import AutoModelForSequenceClassification, AutoTokenizer
-model = AutoModelForSequenceClassification.from_pretrained("mindpadi/emotion_classifier")
-tokenizer = AutoTokenizer.from_pretrained("mindpadi/emotion_classifier")
-inputs = tokenizer("I'm feeling really down", return_tensors="pt")
-outputs = model(**inputs)
-print(outputs.logits)
-```
-### 🌐 Using Inference Endpoints
-```python
-import requests
-url = "https://<your-endpoint>.hf.space"
-headers = {
-    "Authorization": "Bearer <your-token>",
-    "Content-Type": "application/json"
-}
-payload = {"inputs": "What should I do when I'm anxious?"}
-res = requests.post(url, json=payload, headers=headers)
-print(res.json())
-```
-### 🔁 MindPadi Integration
-| File                               | Model Used                            |
-| ---------------------------------- | ------------------------------------- |
-| `app/chatbot/core.py`              | Text generation                       |
-| `app/chatbot/emotion.py`           | `emotion_classifier`                  |
-| `app/chatbot/intent_classifier.py` | `intent_classifier`, `intent_encoder` |
-| `app/utils/embedding_search.py`    | `sentence_transformer`                |
-## 🧩 Ethical Considerations
-* **Supportive, Not Diagnostic**: Not a replacement for therapy
-* **Bias Risk**: Model outputs may contain implicit bias
-* **Data Privacy**: User data must be anonymized
-* **Transparency**: Clearly inform users they're chatting with AI
-## 📬 Contact
-* 📧 Email: [admin@mindpad.com](admin@mindpad.com)
-* 🔗 GitHub: [MindPadi](https://github.com/MindPadi)
-* 🤗 Hugging Face: [https://huggingface.co/mindpadi](https://huggingface.co/mindpadi)
 ## 📄 License
-* **License**: MIT
-* **Version**: 1.0
-* **Last Updated**: May 6, 2025

 license: mit
 language:
   - en
 tags:
+  - intent-classification
+  - emotion-detection
+  - mental-health
+  - lstm
+  - sentence-transformers
+  - sklearn
+pipeline_tag: text-classification
 ---
+# 🧠 MindPadi: Hybrid Classifier Suite
+This repository contains auxiliary models for intent and emotion classification used in the **MindPadi** mental health assistant. These models include rule-based, ML-based, and deep learning classifiers trained to detect emotional states, user intent, and conversational cues.
+## 📦 Files
+| File                          | Description                                              |
+|-------------------------------|----------------------------------------------------------|
+| `intent_clf.joblib`           | scikit-learn pipeline for intent classification (TF-IDF) |
+| `intent_sentence_classifier.pkl` | Sentence-level intent classifier (pickle)                |
+| `lstm_tfidf.h5`               | LSTM model trained on TF-IDF vectors                    |
+| `lstm_bert.h5`                | LSTM model trained on BERT embeddings                   |
+| `tfidf_vectorizer.pkl`        | TF-IDF vectorizer for preprocessing text                |
+| `tfidf_embeddings.pkl`        | Cached TF-IDF embeddings for faster lookup              |
+| `bert_embeddings.npy`         | Precomputed BERT embeddings used in training/testing     |
+| `lstm_accuracy_tfidf.png`     | Evaluation plot (TF-IDF model)                          |
+| `lstm_accuracy_bert.png`      | Evaluation plot (BERT model)                            |
+| `model_configs/`              | JSON configs for training and architecture              |
+## 🎯 Tasks Supported
+- **Intent Classification**: Understand what the user is trying to communicate.
+- **Emotion Detection**: Identify the emotional tone (e.g., sad, angry).
+- **Embedding Generation**: Support vector similarity or hybrid routing.
+## 🔬 Model Overview
+| Model Type     | Framework   | Notes                                 |
+|----------------|-------------|----------------------------------------|
+| LSTM + TF-IDF  | Keras       | Traditional pipeline with good generalization |
+| LSTM + BERT    | Keras       | Handles contextual sentence meanings  |
+| TF-IDF + SVM   | scikit-learn | Lightweight and interpretable intent routing |
+| Sentence Classifier | scikit-learn | Quick rule or decision-tree model for sentence-level labels |
+---
+## 🛠️ Usage Example
+### Intent Prediction (Joblib)
+```python
+from joblib import load
+clf = load("intent_clf.joblib")
+text = ["I feel really anxious today"]
+pred = clf.predict(text)
+print("Intent:", pred[0])
+````
+### LSTM Emotion Prediction
 ```python
+from tensorflow.keras.models import load_model
+import numpy as np
+model = load_model("lstm_bert.h5")
+embeddings = np.load("bert_embeddings.npy")  # assuming aligned with test set
+output = model.predict(embeddings)
+print("Predicted emotion class:", output.argmax(axis=1))
 ```
+## 📊 Evaluation
+| Model               | Accuracy | Dataset Size | Notes                            |
+| ------------------- | -------- | ------------ | -------------------------------- |
+| `lstm_bert.h5`      | \~88%    | 10,000+      | Best for nuanced emotional input |
+| `lstm_tfidf.h5`     | \~83%    | 10,000+      | Lighter, faster                  |
+| `intent_clf.joblib` | \~90%    | 8,000+       | Works well with short queries    |
+Evaluation visualizations:
+* ![](lstm_accuracy_bert.png)
+* ![](lstm_accuracy_tfidf.png)
+## ⚠️ Limitations
+* English only
+* May misclassify ambiguous or sarcastic phrases
+* LSTM models require matching vectorizer or embeddings
+## 🧩 Integration
+These models are invoked in:
+* `app/chatbot/intent_classifier.py`
+* `app/chatbot/emotion.py`
+* `app/utils/embedding_search.py`
+## 🧠 Intended Use
+* Mental health journaling feedback
+* Chatbot-based emotion understanding
+* Offline fallback for heavy transformer models
 ## 📄 License
+MIT License – free for commercial and research use.
+*Last updated: May 6, 2025*