gemma7b-coding-gpt4o-100k

This model is a fine-tuned version of google/gemma-7b on the llama-duo/synth_coding_dataset_dedup dataset. It achieves the following results on the evaluation set:

• Loss: 3.9658

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
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
0.5262	0.9989	470	1.3224
0.4826	2.0	941	1.3435
0.4369	2.9989	1411	1.4787
0.3819	4.0	1882	1.7432
0.3345	4.9989	2352	2.1234
0.2875	6.0	2823	2.5846
0.2319	6.9989	3293	3.1057
0.1968	8.0	3764	3.6609
0.1809	8.9989	4234	3.9400
0.1757	9.9894	4700	3.9658

Framework versions

• PEFT 0.11.1
• Transformers 4.40.1
• Pytorch 2.2.0+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1