Added experiment scripts for gemma-7b

Author: AVS1508

experiments/job-22841975-logs.out

@@ -0,0 +1,3 @@
+Removing miniconda version 22.11.1-1
+Loading miniconda version 22.11.1-1
+python3: can't open file '/home/avsingh_umass_edu/COMPSCI-685-Project/experiments/src/reasoning/main.py': [Errno 2] No such file or directory

experiments/recurring_gemma-7b-optimized.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+#SBATCH -c 4                        # Number of Cores per Task
+#SBATCH --mem=50G                   # Requested Memory
+#SBATCH -p gpu                      # Partition
+#SBATCH -G 1                        # Number of GPUs
+#SBATCH --constraint vram48         # GPU Memory
+#SBATCH -t 12:00:00                 # Job time limit
+#SBATCH -o job-%j-logs.out          # %j = job ID
+
+module load miniconda/22.11.1-1
+conda activate cs685
+
+python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --dataset-size 8792 \
+    --recurring-self-consistency \
+    --num-shots 8 \
+    --max-out-tokens 120 \
+    --gpu-memory-utilization 0.9 \
+    --output-file "results/recurring__gsm8k__gemma-7b__output-optimized.json"

experiments/recurring_gemma-7b-optimized_1.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+#SBATCH -c 4                        # Number of Cores per Task
+#SBATCH --mem=50G                   # Requested Memory
+#SBATCH -p gpu                      # Partition
+#SBATCH -G 1                        # Number of GPUs
+#SBATCH --constraint vram48         # GPU Memory
+#SBATCH -t 12:00:00                 # Job time limit
+#SBATCH -o job-%j-logs.out          # %j = job ID
+
+module load miniconda/22.11.1-1
+conda activate cs685
+
+python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --dataset-size 2198 --dataset-offset 0 \
+    --recurring-self-consistency \
+    --num-shots 8 \
+    --max-out-tokens 120 \
+    --gpu-memory-utilization 0.9 \
+    --output-file "results/recurring__gsm8k__gemma-7b__output-optimized_1.json"

experiments/recurring_gemma-7b-optimized_2.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+#SBATCH -c 4                        # Number of Cores per Task
+#SBATCH --mem=50G                   # Requested Memory
+#SBATCH -p gpu                      # Partition
+#SBATCH -G 1                        # Number of GPUs
+#SBATCH --constraint vram48         # GPU Memory
+#SBATCH -t 12:00:00                 # Job time limit
+#SBATCH -o job-%j-logs.out          # %j = job ID
+
+module load miniconda/22.11.1-1
+conda activate cs685
+
+python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --dataset-size 2198 --dataset-offset 2198 \
+    --recurring-self-consistency \
+    --num-shots 8 \
+    --max-out-tokens 120 \
+    --gpu-memory-utilization 0.9 \
+    --output-file "results/recurring__gsm8k__gemma-7b__output-optimized_2.json"

experiments/recurring_gemma-7b-optimized_3.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+#SBATCH -c 4                        # Number of Cores per Task
+#SBATCH --mem=50G                   # Requested Memory
+#SBATCH -p gpu                      # Partition
+#SBATCH -G 1                        # Number of GPUs
+#SBATCH --constraint vram48         # GPU Memory
+#SBATCH -t 12:00:00                 # Job time limit
+#SBATCH -o job-%j-logs.out          # %j = job ID
+
+module load miniconda/22.11.1-1
+conda activate cs685
+
+python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --dataset-size 2198 --dataset-offset 4396 \
+    --recurring-self-consistency \
+    --num-shots 8 \
+    --max-out-tokens 120 \
+    --gpu-memory-utilization 0.9 \
+    --output-file "results/recurring__gsm8k__gemma-7b__output-optimized_3.json"

experiments/recurring_gemma-7b-optimized_4.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+#SBATCH -c 4                        # Number of Cores per Task
+#SBATCH --mem=50G                   # Requested Memory
+#SBATCH -p gpu                      # Partition
+#SBATCH -G 1                        # Number of GPUs
+#SBATCH --constraint vram48         # GPU Memory
+#SBATCH -t 2:00:00                  # Job time limit
+#SBATCH -o job-%j-logs.out          # %j = job ID
+
+module load miniconda/22.11.1-1
+conda activate cs685
+
+python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --dataset-size 2198 --dataset-offset 6594 \
+    --recurring-self-consistency \
+    --num-shots 8 \
+    --max-out-tokens 120 \
+    --gpu-memory-utilization 0.9 \
+    --output-file "results/recurring__gsm8k__gemma-7b__output-optimized_4.json"

experiments/recurring_gemma-7b.sh

@@ -4,7 +4,7 @@
 #SBATCH -p gpu                      # Partition
 #SBATCH -G 1                        # Number of GPUs
 #SBATCH --constraint vram80         # GPU Memory
-#SBATCH -t 23:59:59                 # Job time limit
+#SBATCH -t 12:00:00                 # Job time limit
 #SBATCH -o job-%j-logs.out          # %j = job ID
 
 module load miniconda/22.11.1-1
@@ -15,4 +15,4 @@ python3 src/reasoning/main.py --model "google/gemma-7b" --dataset "gsm8k" --data
     --num-shots 8 \
     --max-out-tokens 120 \
     --gpu-memory-utilization 0.9 \
-    --output-file "results/recurring__gsm8k__gemma-7b__output.json"
+    --output-file "results/recurring__gsm8k__gemma-7b__output.json"