dty1aaa/codellama-7b-instruct-hf-sft

final_sft_CodeLlama-7b-Instruct-hf

This model is a fine-tuned version of codellama/CodeLlama-7b-Instruct-hf on the efficoder dataset. It achieves the following results on the evaluation set:

• Loss: 0.3331

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: 5e-06
• train_batch_size: 8
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2475	0.3401	50	0.2433
0.2425	0.6803	100	0.2314
0.2147	1.0204	150	0.2318
0.1702	1.3605	200	0.2378
0.1758	1.7007	250	0.2396
0.1438	2.0408	300	0.2643
0.0989	2.3810	350	0.2823
0.0991	2.7211	400	0.2799
0.065	3.0612	450	0.3123
0.0601	3.4014	500	0.3348
0.0645	3.7415	550	0.3324

Framework versions

• Transformers 4.44.2
• Pytorch 2.2.2+cu121
• Datasets 4.8.4
• Tokenizers 0.19.1