Add model documentation

README.md

@@ -0,0 +1,29 @@
+# CSQA GPT2-Large Context-Aware Model
+
+This model is a GPT2-large based model fine-tuned for the CommonsenseQA (CSQA) task with context-aware capabilities.
+
+## Model Architecture
+
+This is a multi-component model that includes:
+- **Encoder Model**: GPT2-large based encoder with adapter layers
+- **Latent Model**: GPT2-large based latent representation model with adapter layers  
+- **Decoder Model**: GPT2-large based decoder with adapter layers
+- **Projection Layers**: Linear projections between encoder-latent and latent-decoder components
+
+## Files Structure
+
+- `encoder.pt` / `encoder_model/`: Encoder component weights and configuration
+- `latent_model.pt` / `latent_model/`: Latent model component weights and configuration  
+- `decoder.pt` / `decoder_model/`: Decoder component weights and configuration
+- `encoder_to_latent_model_proj.pt`: Projection layer from encoder to latent model
+- `latent_model_to_decoder_proj.pt`: Projection layer from latent model to decoder
+- `tokenizer/`: GPT2 tokenizer files
+- `config.json`: Model configuration
+
+## Usage
+
+This model was trained for the CommonsenseQA task and includes specialized components for context-aware reasoning.
+
+## Training
+
+The model was trained in multiple stages on the CommonsenseQA dataset, incorporating context-aware mechanisms to improve reasoning capabilities.