feat: add mamba and dynamic chunking related code and test code

scripts/classification/DC_PatchTST.sh

@@ -0,0 +1,142 @@
+export CUDA_VISIBLE_DEVICES=0
+
+model_name=DC_PatchTST
+
+
+
+# DC_PatchTST specific parameters
+d_model_stage0=64        # Stage 0 dimension (D0)
+depth_enc0=1             # Stage 0 Mamba2 encoder depth  
+depth_enc1=1             # Stage 1 Mamba2 encoder depth
+target_ratio0=0.25       # Target compression ratio for stage 0
+target_ratio1=0.25        # Target compression ratio for stage 1
+
+# EthanolConcentration dataset
+python -u run.py \
+  --task_name classification \
+  --is_training 1 \
+  --root_path ./dataset/EthanolConcentration/ \
+  --model_id EthanolConcentration \
+  --model $model_name \
+  --data UEA \
+  --e_layers 3 \
+  --batch_size 8 \
+  --d_model 128 \
+  --d_ff 256 \
+  --n_heads 8 \
+  --dropout 0.1 \
+  --activation gelu \
+  --des 'DC_PatchTST_Exp' \
+  --itr 1 \
+  --learning_rate 0.0002 \
+  --train_epochs 100 \
+  --patience 10 \
+  --d_model_stage0 $d_model_stage0 \
+  --depth_enc0 $depth_enc0 \
+  --depth_enc1 $depth_enc1 \
+  --target_ratio0 $target_ratio0 \
+  --target_ratio1 $target_ratio1
+
+# FaceDetection dataset  
+python -u run.py \
+  --task_name classification \
+  --is_training 1 \
+  --root_path ./dataset/FaceDetection/ \
+  --model_id FaceDetection \
+  --model $model_name \
+  --data UEA \
+  --e_layers 3 \
+  --batch_size 8 \
+  --d_model 128 \
+  --d_ff 256 \
+  --n_heads 8 \
+  --dropout 0.1 \
+  --activation gelu \
+  --des 'DC_PatchTST_Exp' \
+  --itr 1 \
+  --learning_rate 0.001 \
+  --train_epochs 100 \
+  --patience 10 \
+  --d_model_stage0 $d_model_stage0 \
+  --depth_enc0 $depth_enc0 \
+  --depth_enc1 $depth_enc1 \
+  --target_ratio0 $target_ratio0 \
+  --target_ratio1 $target_ratio1
+
+# Handwriting dataset
+python -u run.py \
+  --task_name classification \
+  --is_training 1 \
+  --root_path ./dataset/Handwriting/ \
+  --model_id Handwriting \
+  --model $model_name \
+  --data UEA \
+  --e_layers 3 \
+  --batch_size 8 \
+  --d_model 128 \
+  --d_ff 256 \
+  --n_heads 8 \
+  --dropout 0.1 \
+  --activation gelu \
+  --des 'DC_PatchTST_Exp' \
+  --itr 1 \
+  --learning_rate 0.001 \
+  --train_epochs 100 \
+  --patience 10 \
+  --d_model_stage0 $d_model_stage0 \
+  --depth_enc0 $depth_enc0 \
+  --depth_enc1 $depth_enc1 \
+  --target_ratio0 $target_ratio0 \
+  --target_ratio1 $target_ratio1
+
+# Heartbeat dataset
+python -u run.py \
+  --task_name classification \
+  --is_training 1 \
+  --root_path ./dataset/Heartbeat/ \
+  --model_id Heartbeat \
+  --model $model_name \
+  --data UEA \
+  --e_layers 3 \
+  --batch_size 8 \
+  --d_model 128 \
+  --d_ff 256 \
+  --n_heads 8 \
+  --dropout 0.1 \
+  --activation gelu \
+  --des 'DC_PatchTST_Exp' \
+  --itr 1 \
+  --learning_rate 0.001 \
+  --train_epochs 100 \
+  --patience 10 \
+  --d_model_stage0 $d_model_stage0 \
+  --depth_enc0 $depth_enc0 \
+  --depth_enc1 $depth_enc1 \
+  --target_ratio0 $target_ratio0 \
+  --target_ratio1 $target_ratio1
+
+# JapaneseVowels dataset
+python -u run.py \
+  --task_name classification \
+  --is_training 1 \
+  --root_path ./dataset/JapaneseVowels/ \
+  --model_id JapaneseVowels \
+  --model $model_name \
+  --data UEA \
+  --e_layers 3 \
+  --batch_size 8 \
+  --d_model 128 \
+  --d_ff 256 \
+  --n_heads 8 \
+  --dropout 0.1 \
+  --activation gelu \
+  --des 'DC_PatchTST_Exp' \
+  --itr 1 \
+  --learning_rate 0.001 \
+  --train_epochs 100 \
+  --patience 10 \
+  --d_model_stage0 $d_model_stage0 \
+  --depth_enc0 $depth_enc0 \
+  --depth_enc1 $depth_enc1 \
+  --target_ratio0 $target_ratio0 \
+  --target_ratio1 $target_ratio1