| config: glm_config.yaml | |
| model: | |
| family: gpt2 | |
| n_dims: 10 | |
| n_embd: 256 | |
| n_head: 8 | |
| n_layer: 12 | |
| n_positions: 101 | |
| out_dir: /home/derixu/Documents/in-context-learning-GLM/glm_weights/08229e17-10ab-46c3-b789-fd94c0d91778 | |
| test_run: false | |
| training: | |
| batch_size: 64 | |
| curriculum: | |
| dims: | |
| end: 10 | |
| inc: 0 | |
| interval: 10000 | |
| start: 10 | |
| points: | |
| end: 40 | |
| inc: 0 | |
| interval: 10000 | |
| start: 40 | |
| data: gaussian | |
| keep_every_steps: 1000 | |
| learning_rate: 0.00025 | |
| num_tasks: null | |
| num_training_examples: null | |
| resume_id: null | |
| save_every_steps: 1000 | |
| task: GLM | |
| task_kwargs: | |
| function_type: logistic | |
| scaling: 0.32 | |
| train_steps: 8000 | |
| wandb: | |
| entity: derryxu | |
| log_every_steps: 100 | |
| name: null | |
| notes: ICL GLM training | |
| project: in-context-training | |