| # 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. |