ADD: updates for training

.vscode/launch.json

@@ -162,12 +162,12 @@
       "request": "launch",
       "program": "${workspaceFolder}/src/train_gpt.py",
       "args": [
-        "--training-steps=500",
+        "--training-steps=50000",
         "--eval_every_n_steps=100",
-        "--log-every-n-steps=1",
-        "--per-device-training-batch-size=1",
-        "--per-device-validation-batch-size=1",
-        "--num-workers=0",
+        "--log-every-n-steps=10",
+        "--per-device-training-batch-size=32",
+        "--per-device-validation-batch-size=32",
+        "--num-workers=32",
         "--num_chunks=32",
         "--chunk_len=500",
         "--chunk_ovlp=50",

scripts/preprocess.sh

@@ -0,0 +1,8 @@
+python src/eeg/preprocess.py \
+    --input_directory data/tuh_eeg \
+    --output_directory data/npy_tuh_eeg \
+    --notch_filter 50 60 \
+    --bandpass_filter 1 48 \
+    --tuh_eeg \
+    --verbose \
+    --parallel

scripts/train.sh

@@ -1 +1,17 @@
-python3 src/train_gpt.py --training-steps=50000 --eval_every_n_steps=1000 --log-every-n-steps=3000 --per-device-training-batch-size=32 --per-device-validation-batch-size=32 --num-workers=16 --num_chunks=32 --chunk_len=256 --chunk_ovlp=32 --num-hidden-layers=6 --num-encoder-layers=6 --run-name='32clen2_embed1024' --training-style='CSM_causal' --embedding-dim=1024 --train-data-path='data/npy_tuh_eeg'
+python src/train_gpt.py \
+    --training-steps=50000 \
+    --eval_every_n_steps=100 \
+    --log-every-n-steps=10 \
+    --per-device-training-batch-size=32 \
+    --per-device-validation-batch-size=32 \
+    --num-workers=32 \
+    --num_chunks=32 \
+    --chunk_len=500 \
+    --chunk_ovlp=50 \
+    --num-hidden-layers=6 \
+    --num-encoder-layers=6 \
+    --run-name=32clen2_embed1024 \
+    --training-style=CSM_causal \
+    --embedding-dim=1024 \
+    --train-data-path=data/npy_tuh_eeg \
+    --verbose=True

scripts/train_parallel.sh

@@ -0,0 +1,18 @@
+python -m torch.distributed.launch --nproc_per_node=2 \
+    src/train_gpt.py \
+    --training-steps=50000 \
+    --eval_every_n_steps=100 \
+    --log-every-n-steps=1 \
+    --per-device-training-batch-size=32 \
+    --per-device-validation-batch-size=32 \
+    --num-workers=16 \
+    --num_chunks=32 \
+    --chunk_len=500 \
+    --chunk_ovlp=50 \
+    --num-hidden-layers=6 \
+    --num-encoder-layers=6 \
+    --run-name=32clen2_embed1024_multi_gpu \
+    --training-style=CSM_causal \
+    --embedding-dim=1024 \
+    --train-data-path=data/npy_tuh_eeg \
+    --verbose=True