AVR/train_configs/qwen3vl_2b_full_sft_all.yaml at main · RunRiotComeOn/AVR · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
### Model Configuration
model_name_or_path: Qwen/Qwen3-VL-2B-Instruct
image_max_pixels: 262144
video_max_pixels: 16384
trust_remote_code: true

### Training Method
stage: sft
do_train: true
finetuning_type: full
freeze_vision_tower: true
freeze_multi_modal_projector: true
freeze_language_model: false
deepspeed: examples/deepspeed/ds_z2_config.json

### Dataset Configuration
# dataset: sft_dataset
dataset: sft_dataset
template: qwen3_vl_nothink
cutoff_len: 2048
overwrite_cache: true
preprocessing_num_workers: 32
dataloader_num_workers: 8

### Output Configuration
output_dir: saves/qwen3vl-2b/full/vision_sr1_1_epoch_no_st
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
save_only_model: false
report_to: tensorboard

### Training Hyperparameters
per_device_train_batch_size: 4
gradient_accumulation_steps: 2
learning_rate: 5.0e-5
num_train_epochs: 1.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000
resume_from_checkpoint: null

### Evaluation (optional - uncomment to enable)
# val_size: 0.1
# per_device_eval_batch_size: 1
# eval_strategy: steps
# eval_steps: 500