updated compression / decompression scripts

.gitattributes

@@ -54,7 +54,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.jpeg filter=lfs diff=lfs merge=lfs -text
 *.webp filter=lfs diff=lfs merge=lfs -text
 # Jsonl/log files
-*.jsonl filter=lfs diff=lfs merge=lfs -text
 *.log filter=lfs diff=lfs merge=lfs -text
-**/*.jsonl filter=lfs diff=lfs merge=lfs -text
-**/*.log filter=lfs diff=lfs merge=lfs -text
+**/*.log filter=lfs diff=lfs merge=lfs -text
+# .tar.gz files
+**/*.tar.gz filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,5 +1,8 @@
 # Skip common hidden files
 .*
 
+# Ignore jsonl files, they are too many/big, so we compress them
+*.jsonl
+
 # Allow back git related files
 !.git*

scripts/build-jsonl-files-from-archive.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+# Check for argument representing the directory to perform the extraction operation in
+if [ -z "$1" ]; then
+    echo "Please provide the directory to perform the extraction operation in"
+    exit 1
+fi
+
+# Go to the directory
+cd "$1"
+
+# Check if there is an existing result-jsonl.tar.gz file in the directory
+if [ ! -f "result-jsonl.tar.gz" ]; then
+    echo "[SKIP] No result-jsonl.tar.gz file found in: $1"
+    exit 0
+fi
+
+# Extract the jsonl files
+echo "[START] Extracting files from result-jsonl.tar.gz into $1"
+tar -xzf result-jsonl.tar.gz
+echo "[DONE] Files extracted from result-jsonl.tar.gz into $1"

scripts/compress-all-jsonl.sh

@@ -15,7 +15,7 @@ echo "==================================================="
 find "$BASE_DIR" -type f -name '*.jsonl' | awk -F/ 'BEGIN{OFS="/"}{$NF=""; print $0}' | xargs -I {} readlink -f {} | sort -u | while read -r DIR
 do
     # Process each directory
-    "$SCRIPT_DIR/generate-result-jsonl-in-folder.sh" "$DIR" &
+    bash "$SCRIPT_DIR/build-jsonl-archive-in-folder.sh" "$DIR" &
 done
 
 # Sleep for 5 second to allow the background processes to start
@@ -23,6 +23,8 @@ sleep 5
 
 # Wait for all background processes to finish
 wait
+wait
+wait
 
 # Sleep for 1 second to allow the background processes to finish / echo logs to flush, etc
 sleep 1

scripts/decompress-all-jsonl.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+# Get the current script directory of the script
+SCRIPT_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )"
+
+# lm-eval-output dir
+PROJECT_DIR="$(dirname "$SCRIPT_DIR")"
+BASE_DIR="$PROJECT_DIR/lm-eval-output"
+
+echo "==================================================="
+echo "Decompressiong all result-jsonl.tar.gz files in $BASE_DIR"
+echo "==================================================="
+
+# Find all directories containing .jsonl files, without duplicates
+find "$BASE_DIR" -type f -name 'result-jsonl.tar.gz' | awk -F/ 'BEGIN{OFS="/"}{$NF=""; print $0}' | xargs -I {} readlink -f {} | sort -u | while read -r DIR
+do
+    # Process each directory
+    bash "$SCRIPT_DIR/build-jsonl-files-from-archive.sh" "$DIR" &
+done
+
+# Sleep for 5 second to allow the background processes to start
+sleep 5
+
+# Wait for all background processes to finish
+wait
+wait
+wait
+
+# Sleep for 1 second to allow the background processes to finish / echo logs to flush, etc
+sleep 1
+
+# Done
+echo "==================================================="
+echo "All jsonl files in $BASE_DIR have been unpacked"
+echo "==================================================="