Ffn sharing

[DylanASHillier]

Research Series: Weight Sharing

• Experiments over forms of weight sharing including FFN-sharing with LoRA (our hopeful/ ours)
• Fun times for all

Adding the experimental code required for the Weight Sharing Experiments.
Will add the prereport once it has been completed

• Fixes the attention to work for different values of hidden dim
• adds a load of configs to run the experiments I want to run
• fixes the base config to not use dropout and log appropriately