【PPMix No.3】support qwen2-vl-2b lora finetune (#847)

support qwen2-vl lora finetue 1.update file 'paddlemix/examples/qwen2_vl/qwen2vl_finetune.py' to support lora finetune 2.add new file 'paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh' 3.update file 'paddlemix/examples/qwen2_vl/README.md' ![image](https://github.com/user-attachments/assets/a4882142-ec5f-4e79-936c-535641c0d29b) --------- Co-authored-by: wcx <[email protected]> Co-authored-by: nifeng <[email protected]>
PaddlePaddle · Dec 2, 2024 · 2ef9368 · 2ef9368
1 parent 9e196f8
commit 2ef9368
Show file tree

Hide file tree

Showing 4 changed files with 139 additions and 3 deletions.
diff --git a/paddlemix/demo_images/qwen2-vl-2b-lora-ft.png b/paddlemix/demo_images/qwen2-vl-2b-lora-ft.png
diff --git a/paddlemix/examples/qwen2_vl/README.md b/paddlemix/examples/qwen2_vl/README.md
@@ -51,15 +51,20 @@ opensource_json.tar需下载解压在playground/目录下，opensource_json 里
 
 ### 4.2 微调命令
 
-注意：此微调训练为全参数微调，冻结视觉编码器而放开LLM训练，2B模型微调训练的显存大小约为30G，7B模型微调训练的显存大小约为75G。
+注意：此微调训练为语言模型微调，冻结视觉编码器而放开LLM训练，2B模型全量微调训练的显存大小约为30G，7B模型全量微调训练的显存大小约为75G。
 
 ```bash
 # 2B
 sh paddlemix/examples/qwen2_vl/shell/basline_2b_bs32_1e8.sh
 
+# 2B lora
+sh paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh
+
 # 7B
 sh paddlemix/examples/qwen2_vl/shell/basline_7b_bs32_1e8.sh
 ```
+注意：微调2b模型的运行示例如下：
+![运行示例](../../demo_images/qwen2-vl-2b-lora-ft.png)
 
 ### 4.3 微调后使用
 

diff --git a/paddlemix/examples/qwen2_vl/qwen2vl_finetune.py b/paddlemix/examples/qwen2_vl/qwen2vl_finetune.py
@@ -20,7 +20,7 @@
 import sys
 import traceback
 from dataclasses import dataclass, field
-from typing import Dict, Optional
+from typing import Dict, Optional, List
 
 import numpy as np
 import paddle
@@ -31,6 +31,7 @@
 from paddlenlp.trainer.trainer import Trainer
 from paddlenlp.trainer.trainer_utils import get_last_checkpoint
 from paddlenlp.transformers import Qwen2Tokenizer
+from paddlenlp.peft import LoRAConfig, LoRAModel
 from PIL import Image, ImageFile, PngImagePlugin, UnidentifiedImageError
 
 from paddlemix.datasets.internvl_dataset import ConcatDataset, WeightedConcatDataset
@@ -125,7 +126,31 @@ class ModelArguments(ProcessorArguments):
         default=0.0,
         metadata={"help": "Set the drop path rate for the ViT model. Default is 0."},
     )
-
+    lora: Optional[bool] = field(
+        default=False,
+        metadata={"help": "Whether or not to use lora to train model."},
+    )    
+    lora_path: Optional[str] = field(
+        default=None, 
+        metadata={"help": "Initialize lora state dict."}
+    )
+    lora_rank: Optional[int] = field(
+        default=128,
+        metadata={"help": "Set the value of rank in lora. Default is 128."},
+    )
+    lora_alpha: Optional[int] = field(
+        default=256,
+        metadata={"help": "Set the value of alpha in lora. Default is 256."},
+    )
+    lora_dropout: Optional[float] = field(
+        default=0.0,
+        metadata={"help": "Set the value of dropout in lora. Default is 0.0."},
+    )
+    lora_target_modules: Optional[str] = field(
+        default=None, 
+        metadata={"help": "Lora target modules."}
+    )
+
 
 @dataclass
 class DataTrainingArguments:
@@ -528,6 +553,26 @@ def _freeze_params(module):
         model.lm_head = model.lm_head.eval()
         _freeze_params(model.model)
         _freeze_params(model.lm_head)
+
+
+    # lora
+    if model_args.lora:
+        if model_args.lora_path is None:
+            target_modules = model_args.lora_target_modules.split(",")
+            lora_config = LoRAConfig(
+                target_modules=target_modules,
+                r=model_args.lora_rank,
+                lora_alpha=model_args.lora_alpha,
+                lora_dropout=model_args.lora_dropout,
+                merge_weights=False,
+                tensor_parallel_degree=training_args.tensor_parallel_degree,
+                dtype=dtype,
+            )
+            model = LoRAModel(model, lora_config)
+        else:
+            model = LoRAModel.from_pretrained(model=model, lora_path=model_args.lora_path)
+        model.mark_only_lora_as_trainable()
+        model.print_trainable_parameters()
 
     print_trainable_params(model)
 

diff --git a/paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh b/paddlemix/examples/qwen2_vl/shell/basline_2b_lora_bs32_1e8.sh
@@ -0,0 +1,86 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#     http://www.apache.org/licenses/LICENSE-2.0
+# 
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+set -x
+
+GPUS=${GPUS:-8}
+BATCH_SIZE=${BATCH_SIZE:-32}
+PER_DEVICE_BATCH_SIZE=${PER_DEVICE_BATCH_SIZE:-1}
+
+GRADIENT_ACC=$((BATCH_SIZE / PER_DEVICE_BATCH_SIZE / GPUS))
+tensor_parallel_degree=${tensor_parallel_degree:-1}
+sharding_parallel_degree=$((GPUS / tensor_parallel_degree))
+
+export PYTHONPATH="${PYTHONPATH}:$(pwd)"
+export MASTER_PORT=34229
+export TF_CPP_MIN_LOG_LEVEL=3
+
+OUTPUT_DIR='work_dirs/basline_330k_2b_bs32_1e8'
+
+if [ ! -d "$OUTPUT_DIR" ]; then
+  mkdir -p "$OUTPUT_DIR"
+fi
+
+TRAINING_MODEL_RESUME="None"
+TRAINER_INSTANCES='127.0.0.1'
+MASTER='127.0.0.1:8080'
+
+meta_path="paddlemix/examples/qwen2_vl/configs/baseline_6data_330k.json"
+
+TRAINING_PYTHON="python -m paddle.distributed.launch --master ${MASTER} --nnodes 1 --nproc_per_node ${GPUS} --rank 0 --ips ${TRAINER_INSTANCES} --run_mode=collective"
+${TRAINING_PYTHON} --log_dir ${OUTPUT_DIR}/paddle_distributed_logs \
+  paddlemix/examples/qwen2_vl/qwen2vl_finetune.py \
+  --do_train \
+  --model_name_or_path "Qwen/Qwen2-VL-2B-Instruct" \
+  --output_dir ${OUTPUT_DIR} \
+  --logging_dir ${OUTPUT_DIR}/logs \
+  --meta_path ${meta_path} \
+  --overwrite_output_dir True \
+  --dataloader_num_workers 8 \
+  --bf16 True \
+  --fp16 False \
+  --fp16_opt_level "O2" \
+  --num_train_epochs 1 \
+  --per_device_train_batch_size ${PER_DEVICE_BATCH_SIZE} \
+  --gradient_accumulation_steps ${GRADIENT_ACC} \
+  --freeze_vit True \
+  --max_seq_length 8192 \
+  --image_resolution 512 \
+  --recompute False \
+  --max_grad_norm 1.0 \
+  --evaluation_strategy "no" \
+  --save_strategy "steps" \
+  --save_steps 1000 \
+  --save_total_limit 1 \
+  --learning_rate 1e-8 \
+  --warmup_ratio 0.1 \
+  --warmup_steps 100 \
+  --weight_decay 0.1 \
+  --optim "adamw" \
+  --lr_scheduler_type "cosine" \
+  --logging_steps 1 \
+  --report_to "visualdl" \
+  --tensor_parallel_degree=${tensor_parallel_degree} \
+  --sharding_parallel_degree=${sharding_parallel_degree} \
+  --pipeline_parallel_degree=1 \
+  --sep_parallel_degree=1 \
+  --sharding="stage1" \
+  --amp_master_grad=1 \
+  --hybrid_parallel_topo_order="sharding_first" \
+  --lora True \
+  --lora_rank=12 \
+  --lora_alpha=256 \
+  --lora_dropout=0.0 \
+  --lora_target_modules="model.layers.*q_proj.*,model.layers.*k_proj.*,model.layers.*v_proj.*,model.layers.*gate_proj.*,model.layers.*up_proj.*,model.layers.*down_proj.*,model.layers.*o_proj.*" \
+  2>&1 | tee -a "${OUTPUT_DIR}/training_log.txt"