kind: nla_model
schema_version: 2
role: ar
stage: sft
base_checkpoint: /home/sparrow/NLAmodel/models/qwen2.5-7b-instruct
d_model: 3584
extraction:
  injection_scale: 150.0
  mse_scale: 59.86651818838306
tokens:
  injection_char: "\u320E"
  injection_token_id: 149705
  injection_left_neighbor_id: 29
  injection_right_neighbor_id: 522
  critic_suffix_ids:
  - 1318
  - 29
  - 366
  - 1708
  - 29
prompt_templates:
  av: 'You are a meticulous AI researcher conducting an important investigation into
    activation vectors from a language model. Your overall task is to describe the
    semantic content of that activation vector.


    We will pass the vector enclosed in <concept> tags into your context. You must
    then produce an explanation for the vector, enclosed within <explanation> tags.
    The explanation consists of 2-3 text snippets describing that vector.


    Here is the vector:


    <concept>{injection_char}</concept>


    Please provide an explanation.'
  ar: 'Summary of the following text: <text>{explanation}</text> <summary>'
critic:
  extraction_layer_index: 20
trained_on:
- /home/sparrow/NLAmodel/experiments/Exp16_affect_av_dataset/runs/production10000_clean_sources/dataset/records.jsonl
parent_checkpoints:
- /home/sparrow/NLAmodel/models/qwen2.5-7b-instruct
created_by: Exp18_affect_ar_sft.train_ar_sft
training:
  optimizer_steps: 209
  learning_rate: 6.12e-06
  global_batch_size: 24
  num_epoch: 1
  loss_type: direction_only_normalized_mse
  mse_scale: 59.86651818838306
  dataset: /home/sparrow/NLAmodel/experiments/Exp16_affect_av_dataset/runs/production10000_clean_sources/dataset/records.jsonl