d_model: 3584
extraction:
injection_scale: null
mse_scale: 59.87
extraction_layer_index: 20
kind: nla_model
prompt_templates:
ar: 'Summary of the following text: {explanation} {injection_char}'
role: ar
schema_version: 2
stage: sl
tokens:
injection_char: "\u320E"
injection_token_id: 149705
training:
best_val_cosine: 0.942993040337707
best_val_mse: 1608.1869257146661
corpus: nla-at-home
epochs: 10
lora_r: 16
lr: 7.0e-05
method: lora_sl
mse_scale: 59.87
n_train: 4753
n_val: 528