meng-lab/llama_3.1_8b_instruct_paradec_humaneval

Llama-3.1-8B-Instruct-sft-5e-3-epoch-100-human-eval-final

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the meng-lab/Llama-3.1-8B-Instruct-humaneval dataset. It achieves the following results on the evaluation set:

• Loss: 5.3754
• Loss Layer 4 Head: 1.6774
• Loss Layer 8 Head: 1.3806
• Loss Layer 12 Head: 1.2795
• Loss Layer 16 Head: 0.6378
• Loss Layer 20 Head: 0.3110
• Loss Layer 24 Head: 0.1844
• Loss Layer 28 Head: 0.0864

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: 0.005
• train_batch_size: 1
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 32
• total_train_batch_size: 128
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Loss Layer 4 Head	Loss Layer 8 Head	Loss Layer 12 Head	Loss Layer 16 Head	Loss Layer 20 Head	Loss Layer 24 Head	Loss Layer 28 Head
7.7477	9.6823	200	7.6952	1.9941	1.7442	1.9609	1.0923	0.4414	0.2459	0.4381
5.8078	19.3646	400	6.4289	1.9090	1.5288	1.4099	0.9812	0.3976	0.2383	0.1448
4.8435	29.0469	600	5.9964	1.8480	1.5236	1.3836	0.6737	0.3976	0.2537	0.1092
4.6084	38.7292	800	6.0069	1.8460	1.7121	1.3111	0.6743	0.3436	0.2146	0.0977
4.0625	48.4115	1000	5.7159	1.8920	1.4329	1.3107	0.6548	0.3220	0.1980	0.0920
3.7565	58.0938	1200	5.4530	1.7095	1.3997	1.2900	0.6451	0.3159	0.1877	0.0897
3.5758	67.7761	1400	5.4088	1.6897	1.3862	1.2843	0.6413	0.3125	0.1860	0.0880
3.5369	77.4584	1600	5.3933	1.6839	1.3837	1.2815	0.6409	0.3124	0.1856	0.0870
3.51	87.1407	1800	5.3780	1.6781	1.3809	1.2799	0.6378	0.3111	0.1843	0.0865
3.4762	96.8230	2000	5.3754	1.6774	1.3806	1.2795	0.6378	0.3110	0.1844	0.0864

Framework versions

• Transformers 4.43.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1