You can deploy this model and start earning money today!

sxj1215/
Qwen2-VL-Synergy

generated_from_trainer

Qwen/Qwen2-VL-7B-Instruct

adapter

other

Views

No views yet

sft_synergy_scienceqalast

This model is a fine-tuned version of Qwen/Qwen2-VL-7B-Instruct on the mmimdb, the memecap, the hateful_memes, the ny_cartoon, the memotion and the scienceqa datasets. It achieves the following results on the evaluation set:

Loss: 0.5355

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 2
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 4
total_train_batch_size: 16
total_eval_batch_size: 2
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.7891	0.1957	500	0.6836
0.7446	0.3914	1000	0.6307
0.7208	0.5870	1500	0.5877
0.6512	0.7827	2000	0.5539
0.6369	0.9784	2500	0.5330
0.47	1.1741	3000	0.5348
0.3866	1.3697	3500	0.5188
0.4721	1.5654	4000	0.5088
0.5444	1.7611	4500	0.4966
0.5069	1.9568	5000	0.4991
0.3624	2.1524	5500	0.5303
0.3805	2.3481	6000	0.5416
0.4058	2.5438	6500	0.5372
0.4088	2.7395	7000	0.5369
0.3336	2.9351	7500	0.5356

Framework versions

PEFT 0.12.0
Transformers 4.45.2
Pytorch 2.1.2+cu121
Datasets 3.1.0
Tokenizers 0.20.3

sxj1215/Qwen2-VL-Synergy

sft_synergy_scienceqalast

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

sxj1215/
Qwen2-VL-Synergy