2_sub_cmd.md

$TASK = "word_onset" $IDX = "idx07"

cd "C:\Users\badri\OneDrive\Documents\EE 675 Neural Learning\Baseline Replication\PopulationTransformer"

python -m data.write_multi_subject_multi_channel "+data_prep=pretrain_multi_subj_multi_chan_template" "++data_prep.task_name=$TASK" "++data_prep.brain_runs=$REPO_DIR/trial_selections/speech_sub1_2_word_onset.json" "++data_prep.electrodes=$REPO_DIR/electrode_selections/clean_laplacian.json" "++data_prep.output_directory=$REPO_DIR/saved_examples/all_test_${TASK}_$IDX" "++data_prep.index_subsample=1.0" "++data_prep.separation_interval=0.5" "+preprocessor=multi_elec_spec_pretrained" "++preprocessor.upstream_ckpt=$STFT_CKPT" "+data=subject_data_template" "++data.cached_transcript_aligns=$CACHED_ALIGNS" "++data.cached_data_array=$CACHED_ARRAYS" "++data.raw_brain_data_dir=$BRAINTREEBANK_DIR" "++data.movie_transcripts_dir=$TRANSCRIPTS_DIR" "++data.rereference=None" "++data.duration=1.0" ` "++data.interval_duration=1.0"

saved_examples/all_test_word_onset_idx07/sub_1/... saved_examples/all_test_word_onset_idx07/sub_2/...

python -m data.make_subject_specific_manifest "+data_prep=subject_specific_manifest" "++data_prep.data_path=$REPO_DIR/saved_examples/all_test_${TASK}$IDX" "++data_prep.subj=sub_1" "++data_prep.out_path=$REPO_DIR/saved_examples/sub_1${TASK}cr$IDX"

python -m data.make_subject_specific_manifest "+data_prep=subject_specific_manifest" "++data_prep.data_path=$REPO_DIR/saved_examples/all_test_${TASK}$IDX" "++data_prep.subj=sub_2" "++data_prep.out_path=$REPO_DIR/saved_examples/sub_2${TASK}cr$IDX"

saved_examples/sub_1_word_onset_cr_idx07/manifest.tsv, labels.tsv, ... saved_examples/sub_2_word_onset_cr_idx07/manifest.tsv, labels.tsv, ...

$TASK = "word_onset" $IDX = "idx07" $SUBJECT = "sub_1" $N = 1 $NAME = "popt_brainbert_stft_${TASK}$IDX" $PRETRAINED_POPT = "$BRAIN_REPO\pretrained_weights\popt_pretrained_weights\pretrained_popt_brainbert_stft.pth" $RESULTS_DIR = "$REPO_DIR\outputs${SUBJECT}${TASK}top${N}${NAME}"

python run_train.py "+exp=multi_elec_feature_extract" "++exp.runner.results_dir=$RESULTS_DIR" "++exp.runner.save_checkpoints=False" "++model.frozen_upstream=True" # linear probe run "+task=pt_feature_extract_coords" "+criterion=pt_feature_extract_coords_criterion" "+preprocessor=empty_preprocessor" "+data=pt_supervised_task_coords" "++data.data_path=$REPO_DIR/saved_examples/${SUBJECT}_${TASK}_cr_$IDX" "++data.saved_data_split=$REPO_DIR/saved_data_splits/${SUBJECT}_${TASK}fine_tuning$IDX" "++data.sub_sample_electrodes=$REPO_DIR/electrode_selections/clean_laplacian.json" "+model=pt_downstream_model" ` "++model.upstream_path=$PRETRAINED_POPT"

If you also want a full fine-tune version, rerun with:

"++model.frozen_upstream=False"

$TASK = "word_onset" $IDX = "idx07" $SUBJECT = "sub_2" $N = 1 $NAME = "popt_brainbert_stft_${TASK}$IDX" $PRETRAINED_POPT = "$BRAIN_REPO\pretrained_weights\popt_pretrained_weights\pretrained_popt_brainbert_stft.pth" $RESULTS_DIR = "$REPO_DIR\outputs${SUBJECT}${TASK}top${N}${NAME}"

python run_train.py "+exp=multi_elec_feature_extract" "++exp.runner.results_dir=$RESULTS_DIR" "++exp.runner.save_checkpoints=False" "++model.frozen_upstream=True" "+task=pt_feature_extract_coords" "+criterion=pt_feature_extract_coords_criterion" "+preprocessor=empty_preprocessor" "+data=pt_supervised_task_coords" "++data.data_path=$REPO_DIR/saved_examples/${SUBJECT}_${TASK}_cr_$IDX" "++data.saved_data_split=$REPO_DIR/saved_data_splits/${SUBJECT}_${TASK}fine_tuning$IDX" "++data.sub_sample_electrodes=$REPO_DIR/electrode_selections/clean_laplacian.json" "+model=pt_downstream_model" ` "++model.upstream_path=$PRETRAINED_POPT"

(Optionally repeat with frozen_upstream=False for a fine-tuned run.)

outputs\sub_1_word_onset_top1_popt_brainbert_stft_word_onset_idx07\results.json outputs\sub_2_word_onset_top1_popt_brainbert_stft_word_onset_idx07\results.json Re-use your existing Python analysis script, just change results_path to point to each of those files in turn, and save the plots or print out:

ROC AUC

best F1 + threshold

precision/recall snapshot