Add files using upload-large-folder tool

.gitattributes

@@ -57,3 +57,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+data_all_eng_slimpj/tokenizer234/tokenizer3/_text_document.idx filter=lfs diff=lfs merge=lfs -text
+data_all_eng_slimpj/tokenizer234/tokenizer4/_text_document.idx filter=lfs diff=lfs merge=lfs -text

create_multi_dataset.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+apt-get update -y
+apt-get install zstd -y
+DIR="/mnt/weka/peacock/idc/wet-data/output/toxic_filtered_without_bloom_new/"
+TRAIN_DIR=$DIR
+OUTPUT_TRAIN_DIR="/mnt/weka/peacock/idc/hineng/data_multi/"
+#rm -rf $OUTPUT_TRAIN_DIR
+#DATASET_NAME="multi"
+#FINAL_DIR="/mnt/weka/peacock/idc/datasets/$DATASET_NAME/"
+#mkdir -p $FINAL_DIR $OUTPUT_TRAIN_DIR
+#langs=($(ls -1 /mnt/weka/peacock/idc/wet-data/output/toxic_filtered_without_bloom_new/2024-10/ | awk '{split($0, a,"_"); print a[1]}' | uniq))
+#for lang in ${langs[@]}; do
+#	files=($(ls $TRAIN_DIR/*/$lang*))
+#	OUTPUT_LANG_DIR=$OUTPUT_TRAIN_DIR/$lang
+#	FINAL_LANG_DIR=$FINAL_DIR/$lang
+#	mkdir -p $OUTPUT_LANG_DIR $FINAL_LANG_DIR
+#	for f in ${files[@]}; do
+#		yr_name=$(echo $f | rev | cut -d/ -f 2 | rev)
+#		new_name=$(echo $f | sed "s/\//_/g")
+#		echo $yr_name
+#		cmd="unzstd $f --stdout > $OUTPUT_LANG_DIR/${yr_name}_$(basename $f).jsonl"
+#		echo $cmd
+#		eval $cmd
+#	done;
+##final_cmd="cat $OUTPUT_LANG_DIR/*.jsonl > $FINAL_LANG_DIR/final_en.jsonl"
+##echo $final_cmd
+##eval $final_cmd
+#done;	
+
+DATASET_NAME="multi"
+FINAL_DIR="/mnt/weka/peacock/idc/datasets/$DATASET_NAME/"
+year_dir=($(ls $TRAIN_DIR))
+langs=("ta" "te" "kn" "ml")
+for l in ${langs[@]}; do
+	for y in ${year_dir[@]}; do
+		echo $y
+		files=($(ls $TRAIN_DIR/$y/$l*))
+		OUTPUT_LANG_DIR=$OUTPUT_TRAIN_DIR/$l
+		FINAL_LANG_DIR=$FINAL_DIR/$l
+		mkdir -p $OUTPUT_LANG_DIR $FINAL_LANG_DIR
+		for f in ${files[@]}; do
+			yr_name=$(echo $f | rev | cut -d/ -f 2 | rev)
+			new_name=$(echo $f | sed "s/\//_/g")
+			echo $yr_name
+			cmd="unzstd $f --stdout > $OUTPUT_LANG_DIR/${yr_name}_$(basename $f).jsonl"
+			echo $cmd
+			eval $cmd
+		done;
+	done;
+done;

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_0.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer0/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaa \
+ --output-prefix $FINAL_DIR/tokenizer0/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_1.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer1/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalab \
+ --output-prefix $FINAL_DIR/tokenizer1/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_10.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer10/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalak \
+ --output-prefix $FINAL_DIR/tokenizer10/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_11.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer11/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalal \
+ --output-prefix $FINAL_DIR/tokenizer11/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_12.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer12/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalam \
+ --output-prefix $FINAL_DIR/tokenizer12/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_13.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer13/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalan \
+ --output-prefix $FINAL_DIR/tokenizer13/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_14.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer14/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalao \
+ --output-prefix $FINAL_DIR/tokenizer14/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_15.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer15/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalap \
+ --output-prefix $FINAL_DIR/tokenizer15/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_16.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer16/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaq \
+ --output-prefix $FINAL_DIR/tokenizer16/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_17.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer17/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalar \
+ --output-prefix $FINAL_DIR/tokenizer17/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_18.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer18/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalas \
+ --output-prefix $FINAL_DIR/tokenizer18/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_19.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer19/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalat \
+ --output-prefix $FINAL_DIR/tokenizer19/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_2.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer2/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalac \
+ --output-prefix $FINAL_DIR/tokenizer2/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_20.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer20/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalau \
+ --output-prefix $FINAL_DIR/tokenizer20/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_21.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer21/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalav \
+ --output-prefix $FINAL_DIR/tokenizer21/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_22.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer22/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaw \
+ --output-prefix $FINAL_DIR/tokenizer22/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_23.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer23/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalax \
+ --output-prefix $FINAL_DIR/tokenizer23/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_24.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer24/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalay \
+ --output-prefix $FINAL_DIR/tokenizer24/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_25.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer25/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaz \
+ --output-prefix $FINAL_DIR/tokenizer25/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_26.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer26/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalba \
+ --output-prefix $FINAL_DIR/tokenizer26/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_27.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer27/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbb \
+ --output-prefix $FINAL_DIR/tokenizer27/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_28.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer28/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbc \
+ --output-prefix $FINAL_DIR/tokenizer28/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_29.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer29/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbd \
+ --output-prefix $FINAL_DIR/tokenizer29/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_3.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer3/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalad \
+ --output-prefix $FINAL_DIR/tokenizer3/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_30.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer30/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbe \
+ --output-prefix $FINAL_DIR/tokenizer30/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_31.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer31/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbf \
+ --output-prefix $FINAL_DIR/tokenizer31/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_32.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer32/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbg \
+ --output-prefix $FINAL_DIR/tokenizer32/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_33.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer33/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbh \
+ --output-prefix $FINAL_DIR/tokenizer33/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_34.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer34/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbi \
+ --output-prefix $FINAL_DIR/tokenizer34/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_35.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer35/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbj \
+ --output-prefix $FINAL_DIR/tokenizer35/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_36.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer36/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbk \
+ --output-prefix $FINAL_DIR/tokenizer36/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_37.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer37/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbl \
+ --output-prefix $FINAL_DIR/tokenizer37/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_38.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer38/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbm \
+ --output-prefix $FINAL_DIR/tokenizer38/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_39.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer39/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalbn \
+ --output-prefix $FINAL_DIR/tokenizer39/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_4.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer4/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalae \
+ --output-prefix $FINAL_DIR/tokenizer4/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_5.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer5/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaf \
+ --output-prefix $FINAL_DIR/tokenizer5/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_6.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer6/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalag \
+ --output-prefix $FINAL_DIR/tokenizer6/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_7.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer7/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalah \
+ --output-prefix $FINAL_DIR/tokenizer7/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_8.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer8/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalai \
+ --output-prefix $FINAL_DIR/tokenizer8/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_9.sh

@@ -0,0 +1,25 @@
+#!/bin/bash
+set -m
+
+echo "above deepspeed"
+FINAL_DIR=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english/
+#TOKENIZER=facebook/nllb-200-distilled-600M
+#TOKENIZER_TYPE=HuggingFaceTokenizer
+TOKENIZER_PATH=/mnt/weka/peacock/tokenization/trained-tokenizer/enhiben_50k
+TOKENIZER=$TOKENIZER_PATH/all.model
+VOCAB_FILE=$TOKENIZER_PATH/all.vocab
+TOKENIZER_TYPE=SentencePieceTokenizer
+mkdir -p $FINAL_DIR/tokenizer9/
+cd /mnt/weka/peacock/experiments/llama/Megatron-DeepSpeed
+echo "inside deepspeed"
+pwd
+python3 tools/preprocess_data.py \
+ --input /mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/split40//finalaj \
+ --output-prefix $FINAL_DIR/tokenizer9/ \
+  --tokenizer-model $TOKENIZER \
+  --vocab-file $VOCAB_FILE  \
+    --dataset-impl mmap --tokenizer-type $TOKENIZER_TYPE\
+    --append-eod --workers 8 # --partitions 16 #--chunk-size 50   
+
+  # --merge-file $MERGES_FILE \
+

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_0.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-0
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-0
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-0-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_0.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_1.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-1
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-1
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-1-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_1.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_10.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-10
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-10
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-10-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_10.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_12.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-12
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-12
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-12-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_12.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_13.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-13
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-13
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-13-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_13.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock

data_all_eng_slimpj/tok_english-5/tok_jobs/tokenizer_14.yaml

@@ -0,0 +1,104 @@
+apiVersion: kubeflow.org/v2beta1
+kind: MPIJob
+metadata:
+  name: peacock-tokenizer-english-5-14
+  namespace: peacock
+  labels:
+    app: peacock-tokenizer-english-5-14
+spec:
+  slotsPerWorker: 8
+  runPolicy:
+    backoffLimit: 1
+    cleanPodPolicy: Running
+  mpiReplicaSpecs:
+    Launcher:
+      replicas: 1
+      template:
+        spec:
+          hostIPC: true
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-tokenizer-english-5-14-launcher
+              imagePullPolicy: IfNotPresent
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock
+              command: ["/bin/bash", "-c"]
+              args:
+                - >-
+                  
+                  export SYNAPSE_VERSION="1.15.1";
+                  export WORKER_DIR="/mnt/weka/peacock/experiments/llama";
+
+                  export MEGATRON_SETUP_CMD="$WORKER_DIR/launch/setup.sh"
+                  export TOKENIZER_CMD=/mnt/weka/peacock/idc/hineng/data_all_eng_slimpj/tok_english-5/tok_files/tokenizer_14.sh
+                  HOSTSFILE=$OMPI_MCA_orte_default_hostfile;
+                  echo "HOSTSFILE=$HOSTSFILE";
+                  MASTER_ADDR="$(head -n 1 $HOSTSFILE | sed -n s/[[:space:]]slots.*//p)";
+                  NUM_NODES=$(wc -l < $HOSTSFILE);
+                  CARDS_PER_NODE=8;
+                  N_CARDS=$((NUM_NODES*CARDS_PER_NODE));
+                  echo "MPI_ROOT=$MPI_ROOT";
+                  echo "N_CARDS=$N_CARDS";
+                  echo "MASTER_ADDR=$MASTER_ADDR";
+                  sleep 20;
+
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $MEGATRON_SETUP_CMD;
+
+                  mpirun -np $N_CARDS  -npernode 8 \
+                    --tag-output \
+                    --allow-run-as-root \
+                    --prefix $MPI_ROOT \
+                    -x WORKER_DIR=$WORKER_DIR \
+                    -x SYNAPSE_VERSION=$SYNAPSE_VERSION $TOKENIZER_CMD;
+                    
+                   
+    Worker:
+      replicas: 1
+      template:
+        spec:
+          volumes:
+            - name: work-dir
+              persistentVolumeClaim:
+                claimName: peacock-fs-pvc
+          tolerations:
+            - key: "habana.ai/gaudi"
+              operator: "Exists"
+              effect: "NoSchedule"
+            - key: "k8s/namespace"
+              operator: "Equal"
+              value: "peacock"
+              effect: "NoSchedule"
+          hostIPC: true
+          containers:
+            - image: vault.habana.ai/gaudi-docker/1.15.1/ubuntu22.04/habanalabs/pytorch-installer-2.2.0:latest
+              name: peacock-llama-worker
+              imagePullPolicy: IfNotPresent
+              securityContext:
+                capabilities:
+                  add:
+                    - SYSLOG
+              resources:
+                limits:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+                requests:
+                  habana.ai/gaudi: 8
+                  hugepages-2Mi: 300Gi
+                  memory: 700Gi
+                  cpu: 150
+              volumeMounts:
+                - name: work-dir
+                  mountPath: /mnt/weka/peacock