From a42fec8b027ca701f8f568ca8ef95299b8151f05 Mon Sep 17 00:00:00 2001
From: Tejas Makode <114249576+TejMakode1523@users.noreply.github.com>
Date: Tue, 6 Aug 2024 10:53:48 +0530
Subject: [PATCH 1/2] Add files via upload

---
 finetune/finetune_lora_int4.sh | 62 ++++++++++++++++++++++++++++++++++
 1 file changed, 62 insertions(+)
 create mode 100644 finetune/finetune_lora_int4.sh

diff --git a/finetune/finetune_lora_int4.sh b/finetune/finetune_lora_int4.sh
new file mode 100644
index 0000000..2ff0fe7
--- /dev/null
+++ b/finetune/finetune_lora_int4.sh
@@ -0,0 +1,62 @@
+#!/bin/bash
+
+GPUS_PER_NODE=1
+NNODES=1
+NODE_RANK=0
+MASTER_ADDR=localhost
+MASTER_PORT=6001
+
+MODEL="openbmb/MiniCPM-Llama3-V-2_5-int4"
+DATA="/home/stg-ops/Tejas/MiniCPM-V/train_dataset/dataset.json"
+EVAL_DATA="/home/stg-ops/Tejas/MiniCPM-V/test_dataset/dataset.json"
+LLM_TYPE="llama3"
+
+DISTRIBUTED_ARGS="
+    --nproc_per_node $GPUS_PER_NODE \
+    --nnodes $NNODES \
+    --node_rank $NODE_RANK \
+    --master_addr $MASTER_ADDR \
+    --master_port $MASTER_PORT
+"
+export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:128
+torchrun $DISTRIBUTED_ARGS finetune.py  \
+    --model_name_or_path $MODEL \
+    --llm_type $LLM_TYPE \
+    --data_path $DATA \
+    --eval_data_path $EVAL_DATA \
+    --remove_unused_columns false \
+    --label_names "labels" \
+    --prediction_loss_only false \
+    --bf16 true \
+    --bf16_full_eval true \
+    --fp16 false \
+    --fp16_full_eval false \
+    --do_train \
+    --do_eval \
+    --tune_vision true \
+    --tune_llm false \
+    --use_lora true \
+    --lora_target_modules "llm\..*layers\.\d+\.self_attn\.(q_proj|k_proj)" \
+    --model_max_length 2048 \
+    --max_slice_nums 9 \
+    --max_steps 1000 \
+    --eval_steps 100 \
+    --output_dir output/output_minicpmv2_lora \
+    --logging_dir output/output_minicpmv2_lora \
+    --logging_strategy "steps" \
+    --per_device_train_batch_size 1 \
+    --per_device_eval_batch_size 1 \
+    --gradient_accumulation_steps 4 \
+    --evaluation_strategy "steps" \
+    --save_strategy "steps" \
+    --save_steps 1000 \
+    --save_total_limit 10 \
+    --learning_rate 1e-6 \
+    --weight_decay 0.1 \
+    --adam_beta2 0.95 \
+    --warmup_ratio 0.01 \
+    --lr_scheduler_type "cosine" \
+    --logging_steps 1 \
+    --gradient_checkpointing true \
+    --deepspeed ds_config_zero2.json \
+    --report_to "tensorboard"

From 0527651ced784d08a47eb4be1d5ee5c8af90b474 Mon Sep 17 00:00:00 2001
From: Tejas Makode <114249576+TejMakode1523@users.noreply.github.com>
Date: Tue, 6 Aug 2024 10:56:03 +0530
Subject: [PATCH 2/2] Update finetune_lora_int4.sh

---
 finetune/finetune_lora_int4.sh | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/finetune/finetune_lora_int4.sh b/finetune/finetune_lora_int4.sh
index 2ff0fe7..9aecb05 100644
--- a/finetune/finetune_lora_int4.sh
+++ b/finetune/finetune_lora_int4.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-
+#Fine tuning of MiniCPM-Llama3-V-2_5-int4 on single GPU
 GPUS_PER_NODE=1
 NNODES=1
 NODE_RANK=0
@@ -7,8 +7,8 @@ MASTER_ADDR=localhost
 MASTER_PORT=6001
 
 MODEL="openbmb/MiniCPM-Llama3-V-2_5-int4"
-DATA="/home/stg-ops/Tejas/MiniCPM-V/train_dataset/dataset.json"
-EVAL_DATA="/home/stg-ops/Tejas/MiniCPM-V/test_dataset/dataset.json"
+DATA="" #Path to json file
+EVAL_DATA=""#Path to json file
 LLM_TYPE="llama3"
 
 DISTRIBUTED_ARGS="