fix single-gpu qlora, and add profiling

finetune/finetune_ds.sh

@@ -41,7 +41,7 @@ torchrun $DISTRIBUTED_ARGS finetune.py \
     --lr_scheduler_type "cosine" \
     --logging_steps 1 \
     --report_to "none" \
-    --model_max_length 2048 \
+    --model_max_length 512 \
     --gradient_checkpointing True \
     --lazy_preprocess True \
     --deepspeed finetune/ds_config_zero3.json

finetune/finetune_lora_ds.sh

@@ -41,7 +41,7 @@ torchrun $DISTRIBUTED_ARGS finetune.py \
     --lr_scheduler_type "cosine" \
     --logging_steps 1 \
     --report_to "none" \
-    --model_max_length 2048 \
+    --model_max_length 512 \
     --lazy_preprocess True \
     --use_lora \
     --gradient_checkpointing \

finetune/finetune_lora_single_gpu.sh

@@ -30,7 +30,7 @@ python finetune.py \
   --lr_scheduler_type "cosine" \
   --logging_steps 1 \
   --report_to "none" \
-  --model_max_length 2048 \
+  --model_max_length 512 \
   --lazy_preprocess True \
   --gradient_checkpointing \
   --use_lora

finetune/finetune_qlora_ds.sh

@@ -42,7 +42,7 @@ torchrun $DISTRIBUTED_ARGS finetune.py \
     --lr_scheduler_type "cosine" \
     --logging_steps 1 \
     --report_to "none" \
-    --model_max_length 2048 \
+    --model_max_length 512 \
     --lazy_preprocess True \
     --use_lora \
     --q_lora \

finetune/finetune_qlora_single_gpu.sh

@@ -0,0 +1,38 @@
+#!/bin/bash
+export CUDA_DEVICE_MAX_CONNECTIONS=1
+DIR=`pwd`
+
+MODEL="Qwen/Qwen-7B-Chat-Int4" # Set the path if you do not want to load from huggingface directly
+# ATTENTION: specify the path to your training data, which should be a json file consisting of a list of conversations.
+# See the section for finetuning in README for more information.
+DATA="path_to_data"
+
+export CUDA_VISIBLE_DEVICES=0
+
+# Remember to use --fp16 instead of --bf16 due to autogptq
+python finetune.py \
+  --model_name_or_path $MODEL \
+  --data_path $DATA \
+  --fp16 True \
+  --output_dir output_qwen \
+  --num_train_epochs 5 \
+  --per_device_train_batch_size 2 \
+  --per_device_eval_batch_size 1 \
+  --gradient_accumulation_steps 8 \
+  --evaluation_strategy "no" \
+  --save_strategy "steps" \
+  --save_steps 1000 \
+  --save_total_limit 10 \
+  --learning_rate 1e-5 \
+  --weight_decay 0.1 \
+  --adam_beta2 0.95 \
+  --warmup_ratio 0.01 \
+  --lr_scheduler_type "cosine" \
+  --logging_steps 1 \
+  --report_to "none" \
+  --model_max_length 512 \
+  --lazy_preprocess True \
+  --gradient_checkpointing \
+  --use_lora \
+  --q_lora \
+  --deepspeed finetune/ds_config_zero2.json