以下配置文件，怎样修改，可以让显存占用最小，从而8G显存不会爆显存？

[nissansz]

Global:
debug: false
use_gpu: true
epoch_num: 200
log_smooth_window: 20
print_batch_step: 10
save_model_dir: ./output/rec_svtrv2_ch
save_epoch_step: 10
eval_batch_step: [0, 1000]
cal_metric_during_train: false
pretrained_model:
checkpoints:
save_inference_dir:
use_visualdl: false
infer_img: doc/imgs_words/ch/word_1.jpg
character_dict_path: ppocr/utils/ppocr_keys_v1.txt
max_text_length: &max_text_length 25
infer_mode: false
use_space_char: true
#distributed: true
distributed: false
save_res_path: ./output/rec/predicts_svrtv2.txt

Optimizer:
name: AdamW
beta1: 0.9
beta2: 0.999
epsilon: 1.e-8
weight_decay: 0.05
no_weight_decay_name: norm
one_dim_param_no_weight_decay: True
lr:
name: Cosine
learning_rate: 0.001 # 8gpus 192bs
warmup_epoch: 5

Architecture:
model_type: rec
algorithm: SVTR_HGNet
Transform:
Backbone:
name: SVTRv2
use_pos_embed: False
dims: [128, 256, 384]
depths: [6, 6, 6]
num_heads: [4, 8, 12]
mixer: [['Conv','Conv','Conv','Conv','Conv','Conv'],['Conv','Conv','Global','Global','Global','Global'],['Global','Global','Global','Global','Global','Global']]
local_k: [[5, 5], [5, 5], [-1, -1]]
sub_k: [[2, 1], [2, 1], [-1, -1]]
last_stage: False
use_pool: True
Head:
name: MultiHead
head_list:
- CTCHead:
Neck:
name: svtr
dims: 256
depth: 2
hidden_dims: 256
kernel_size: [1, 3]
use_guide: True
Head:
fc_decay: 0.00001
- NRTRHead:
nrtr_dim: 384
max_text_length: *max_text_length
num_decoder_layers: 2

Loss:
name: MultiLoss
loss_config_list:
- CTCLoss:
- NRTRLoss:

PostProcess:
name: CTCLabelDecode

Metric:
name: RecMetric
main_indicator: acc

Train:
dataset:
name: MultiScaleDataSet
ds_width: false
data_dir: ./train_data/
ext_op_transform_idx: 1
label_file_list:
- ./train_data/train_list.txt
transforms:
- DecodeImage:
img_mode: BGR
channel_first: false
- RecAug:
- MultiLabelEncode:
gtc_encode: NRTRLabelEncode
- KeepKeys:
keep_keys:
- image
- label_ctc
- label_gtc
- length
- valid_ratio
sampler:
name: MultiScaleSampler
#scales: [[320, 32], [320, 48], [320, 64]]
scales: [[320, 32], [320, 48], [320, 16]]
first_bs: &bs 2
fix_bs: false
divided_factor: [8, 16] # w, h
is_training: True
loader:
shuffle: true
batch_size_per_card: *bs
drop_last: true
num_workers: 8
Eval:
dataset:
name: SimpleDataSet
data_dir: ./train_data
label_file_list:
- ./train_data/val_list.txt
transforms:
- DecodeImage:
img_mode: BGR
channel_first: false
- MultiLabelEncode:
gtc_encode: NRTRLabelEncode
- RecResizeImg:
image_shape: [3, 48, 320]
- KeepKeys:
keep_keys:
- image
- label_ctc
- label_gtc
- length
- valid_ratio
loader:
shuffle: false
drop_last: false
#batch_size_per_card: 128
batch_size_per_card: 2
num_workers: 4

[GreatV]

为了减少显存占用，让模型能在8G显存上运行，可以从以下几个方面修改配置文件：

1. 减小批次大小(batch size)：

Train:
  loader:
    batch_size_per_card: 2  # 从原来的64降到2
Eval:
  loader: 
    batch_size_per_card: 2  # 从原来的128降到2

2. 缩小模型结构：

Architecture:
  Backbone:
    name: SVTRv2
    dims: [64, 128, 192]  # 降低维度 [128, 256, 384] -> [64, 128, 192]
    depths: [3, 3, 3]     # 减少层数 [6, 6, 6] -> [3, 3, 3]
    num_heads: [2, 4, 6]  # 减少注意力头数 [4, 8, 12] -> [2, 4, 6]

3. 减小输入图像尺寸：

Train:
  sampler:
    scales: [[160, 32], [160, 48], [160, 16]]  # 从320降到160

Eval:
  dataset:
    transforms:
      - RecResizeImg:
          image_shape: [3, 48, 160]  # 从320降到160

4. 其他优化：

Global:
  max_text_length: 15     # 从25降到15
  save_epoch_step: 20     # 降低保存频率以节省磁盘空间
  print_batch_step: 20    # 降低打印频率

Architecture:
  Head:
    - CTCHead:
        Neck:
          dims: 128       # 从256降到128
          hidden_dims: 128 # 从256降到128
    - NRTRHead:
          nrtr_dim: 192   # 从384降到192

5. 减少工作进程数：

Train:
  loader:
    num_workers: 2   # 从8降到2
Eval:
  loader:
    num_workers: 2   # 从4降到2

这些修改主要通过以下方式减少显存占用：

• 减小batch size
• 缩小模型结构(降低维度、减少层数、减少注意力头数)
• 减小输入图像尺寸
• 减少最大文本长度
• 降低中间特征维度

注意这些修改可能会影响模型的性能，你可以根据实际需求调整这些参数，在显存占用和模型性能之间找到平衡点。

上述修改组合使用后，应该可以显著降低显存占用，使模型能在8G显存上运行。建议先应用这些修改，然后根据实际运行情况进行微调。

---

Response generated by feifei-bot | claude-3-5-sonnet-latest