qwen2-7b-instruct-trl-sft-ChartQA

This model is a fine-tuned version of Qwen/Qwen2-VL-2B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2406

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
2.8769	0.1130	10	2.3660
2.0843	0.2260	20	1.5363
1.1442	0.3390	30	0.6542
0.5756	0.4520	40	0.4312
0.4821	0.5650	50	0.4034
0.4743	0.6780	60	0.3902
0.4445	0.7910	70	0.3732
0.4353	0.9040	80	0.3488
0.3692	1.0113	90	0.2830
0.3362	1.1243	100	0.2695
0.3219	1.2373	110	0.2662
0.3257	1.3503	120	0.2613
0.3012	1.4633	130	0.2603
0.3132	1.5763	140	0.2568
0.3054	1.6893	150	0.2548
0.3119	1.8023	160	0.2528
0.2956	1.9153	170	0.2507
0.2989	2.0226	180	0.2500
0.3087	2.1356	190	0.2476
0.2943	2.2486	200	0.2481
0.3064	2.3616	210	0.2453
0.2896	2.4746	220	0.2438
0.2825	2.5876	230	0.2435
0.2817	2.7006	240	0.2400
0.2758	2.8136	250	0.2401
0.2775	2.9266	260	0.2406

Framework versions

• PEFT 0.15.2
• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.1