diff --git a/pokedex_EfficientNetV2M.py b/Python/pokedex_EfficientNetV2M.py
similarity index 67%
rename from pokedex_EfficientNetV2M.py
rename to Python/pokedex_EfficientNetV2M.py
index b9fd017804d02d7508a44a62701f498c71c63386..d6207c1b9ae6edd096a3059c443e8a2f4d8ffd12 100644
--- a/pokedex_EfficientNetV2M.py
+++ b/Python/pokedex_EfficientNetV2M.py
@@ -16,32 +16,8 @@ num_classes = 151
base_batch_size = 32
base_lr = 1e-3
-# --- Auto-Tune Batch Size ---
-def find_max_batch_size(data_dir, image_size, candidate_sizes=[256, 128, 64, 32, 16]):
- print("Tuning batch size...")
- for bs in candidate_sizes:
- try:
- print(f"Trying global batch size {bs}...")
- ds = keras.utils.image_dataset_from_directory(
- data_dir,
- labels="inferred",
- label_mode="int",
- image_size=image_size,
- batch_size=bs,
- shuffle=True
- )
- for batch in ds.take(1): # Try to load one batch
- tf.print("✓ Batch size", bs, "works")
- del ds
- gc.collect()
- tf.keras.backend.clear_session()
- return bs
- except tf.errors.ResourceExhaustedError:
- print(f"✗ Batch size {bs} too large.")
- raise RuntimeError("No suitable batch size found.")
-
-global_batch_size = find_max_batch_size(data_dir, image_size)
-scaled_lr = base_lr * (global_batch_size / base_batch_size)
+global_batch_size = 32
+scaled_lr = min(base_lr * (global_batch_size / base_batch_size), 1e-3)
# --- Load Dataset ---
full_ds = keras.utils.image_dataset_from_directory(
@@ -107,7 +83,8 @@ with strategy.scope():
# --- Train ---
callbacks = [
- keras.callbacks.ModelCheckpoint("EfficientNetV2M/save_at_{epoch}.keras")
+ keras.callbacks.ModelCheckpoint("/home/users/d/divia/EfficientNetV2M/save_at_{epoch}.keras"),
+ keras.callbacks.EarlyStopping(monitor="val_loss", patience=3, restore_best_weights=True)
]
model.fit(
diff --git a/pokedex_ResNet50.py b/Python/pokedex_ResNet50.py
similarity index 66%
rename from pokedex_ResNet50.py
rename to Python/pokedex_ResNet50.py
index 9c16244c427f887056aa072ed2725305ec3f2ffc..9659d76a850b341c54a251ea529a35c3cfc9484d 100644
--- a/pokedex_ResNet50.py
+++ b/Python/pokedex_ResNet50.py
@@ -16,32 +16,8 @@ num_classes = 151
base_batch_size = 32
base_lr = 1e-3
-# --- Auto-Tune Batch Size ---
-def find_max_batch_size(data_dir, image_size, candidate_sizes=[256, 128, 64, 32, 16]):
- print("Tuning batch size...")
- for bs in candidate_sizes:
- try:
- print(f"Trying global batch size {bs}...")
- ds = keras.utils.image_dataset_from_directory(
- data_dir,
- labels="inferred",
- label_mode="int",
- image_size=image_size,
- batch_size=bs,
- shuffle=True
- )
- for batch in ds.take(1): # Try to load one batch
- tf.print("✓ Batch size", bs, "works")
- del ds
- gc.collect()
- tf.keras.backend.clear_session()
- return bs
- except tf.errors.ResourceExhaustedError:
- print(f"✗ Batch size {bs} too large.")
- raise RuntimeError("No suitable batch size found.")
-
-global_batch_size = find_max_batch_size(data_dir, image_size)
-scaled_lr = base_lr * (global_batch_size / base_batch_size)
+global_batch_size = 32
+scaled_lr = min(base_lr * (global_batch_size / base_batch_size), 1e-3)
# --- Load Dataset ---
full_ds = keras.utils.image_dataset_from_directory(
@@ -107,12 +83,13 @@ with strategy.scope():
# --- Train ---
callbacks = [
- keras.callbacks.ModelCheckpoint("ResNet50/save_at_{epoch}.keras")
+ keras.callbacks.ModelCheckpoint("/home/users/d/divia/ResNet50/save_at_{epoch}.keras"),
+ keras.callbacks.EarlyStopping(monitor="val_loss", patience=2, restore_best_weights=True)
]
model.fit(
train_ds,
validation_data=val_ds,
- epochs=10,
+ epochs=15,
callbacks=callbacks
)
\ No newline at end of file
diff --git a/models/efficientv2m.keras b/models/efficientv2m.keras
new file mode 100644
index 0000000000000000000000000000000000000000..1c7275d712f61cf9369aeda45df72fc9bd05ce98
Binary files /dev/null and b/models/efficientv2m.keras differ
diff --git a/models/resnet50.keras b/models/resnet50.keras
new file mode 100644
index 0000000000000000000000000000000000000000..ca1c556508e3ffa175c3d3a81015954eaef6fad8
Binary files /dev/null and b/models/resnet50.keras differ
diff --git a/simple_xception.keras b/models/simple_xception.keras
similarity index 100%
rename from simple_xception.keras
rename to models/simple_xception.keras
diff --git a/slurm/EfficientNetV2M_16335505.out b/slurm/EfficientNetV2M_16335505.out
new file mode 100644
index 0000000000000000000000000000000000000000..70b7e0d14ebfaf1fbe199a83208f3dda13ea7067
--- /dev/null
+++ b/slurm/EfficientNetV2M_16335505.out
@@ -0,0 +1,97 @@
+2025-04-02 09:24:18.434290: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA
+To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
+2025-04-02 09:24:26.222632: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.222943: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.500324: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.500659: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.501020: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.501261: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.506807: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: SSE4.1 SSE4.2 AVX AVX2 FMA
+To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
+2025-04-02 09:24:26.678433: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.678718: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.679091: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.679427: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.679750: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:26.679999: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.142087: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.142406: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.144113: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.144367: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.144608: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.144853: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1613] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 79383 MB memory: -> device: 0, name: NVIDIA A100 80GB PCIe, pci bus id: 0000:41:00.0, compute capability: 8.0
+2025-04-02 09:24:28.145150: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:981] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
+2025-04-02 09:24:28.145375: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1613] Created device /job:localhost/replica:0/task:0/device:GPU:1 with 79383 MB memory: -> device: 1, name: NVIDIA A100 80GB PCIe, pci bus id: 0000:81:00.0, compute capability: 8.0
+WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting RngReadAndSkip cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting Bitcast cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting StatelessRandomUniformV2 cause there is no registered converter for this op.
+WARNING:tensorflow:Using a while_loop for converting ImageProjectiveTransformV3 cause there is no registered converter for this op.
+2025-04-02 09:25:00.926897: W tensorflow/core/grappler/optimizers/data/auto_shard.cc:784] AUTO sharding policy will apply DATA sharding policy as it failed to apply FILE sharding policy because of the following reason: Found an unshardable source dataset: name: "TensorSliceDataset/_1"
+op: "TensorSliceDataset"
+input: "Placeholder/_0"
+attr {
+ key: "Toutput_types"
+ value {
+ list {
+ type: DT_STRING
+ }
+ }
+}
+attr {
+ key: "_cardinality"
+ value {
+ i: 25511
+ }
+}
+attr {
+ key: "is_files"
+ value {
+ b: false
+ }
+}
+attr {
+ key: "metadata"
+ value {
+ s: "\n\024TensorSliceDataset:0"
+ }
+}
+attr {
+ key: "output_shapes"
+ value {
+ list {
+ shape {
+ }
+ }
+ }
+}
+attr {
+ key: "replicate_on_split"
+ value {
+ b: false
+ }
+}
+experimental_type {
+ type_id: TFT_PRODUCT
+ args {
+ type_id: TFT_DATASET
+ args {
+ type_id: TFT_PRODUCT
+ args {
+ type_id: TFT_TENSOR
+ args {
+ type_id: TFT_STRING
+ }
+ }
+ }
+ }
+}
+
+Number of GPUs: 2
+Found 25511 files belonging to 151 classes.
+Epoch 1/10
diff --git a/slurm/train_EfficientNetV2M.sh b/slurm/train_EfficientNetV2M.sh
new file mode 100644
index 0000000000000000000000000000000000000000..207757225027be1444ee7604e80cbaedfc038082
--- /dev/null
+++ b/slurm/train_EfficientNetV2M.sh
@@ -0,0 +1,17 @@
+#!/bin/sh
+#SBATCH --job-name=keras_cnn_efficientnetv2m
+#SBATCH --output=EfficientNetV2M_%j.out
+#SBATCH --partition=shared-gpu
+#SBATCH --gres=gpu:2,VramPerGpu:80G
+#SBATCH --cpus-per-task=2
+#SBATCH --mem=16G
+#SBATCH --time=01:00:00
+#SBATCH --mail-type=FAIL
+
+# Load modules
+module purge
+module load GCC/11.3.0 OpenMPI/4.1.4 TensorFlow/2.11.0-CUDA-11.7.0
+module load cuDNN/8.4.1.50-CUDA-11.7.0
+
+# Run your script
+srun python ../Python/pokedex_EfficientNetV2M.py
diff --git a/slurm/train_ResNet50.sh b/slurm/train_ResNet50.sh
new file mode 100644
index 0000000000000000000000000000000000000000..f545644d058abf1d2738d181933c7703c73c06ea
--- /dev/null
+++ b/slurm/train_ResNet50.sh
@@ -0,0 +1,17 @@
+#!/bin/sh
+#SBATCH --job-name=keras_cnn_resnet50
+#SBATCH --output=ResNet50_%j.out
+#SBATCH --partition=shared-gpu
+#SBATCH --gres=gpu:2,VramPerGpu:40G
+#SBATCH --cpus-per-task=2
+#SBATCH --mem=16G
+#SBATCH --time=01:00:00
+#SBATCH --mail-type=FAIL
+
+# Load modules
+module purge
+module load GCC/11.3.0 OpenMPI/4.1.4 TensorFlow/2.11.0-CUDA-11.7.0
+module load cuDNN/8.4.1.50-CUDA-11.7.0
+
+# Run your script
+srun python ../Python/pokedex_ResNet50.py