dalyaff/phi2-viggo-finetune

phi-2

This model is a fine-tuned version of microsoftl on the GEM/viggo dataset. It achieves the following results on the evaluation set:

• Loss: 0.2330

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: 2.5e-05
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss
1.917	0.04	50	1.4649
0.7037	0.08	100	0.4905
0.4209	0.12	150	0.3564
0.3534	0.16	200	0.3127
0.311	0.2	250	0.2940
0.2944	0.24	300	0.2798
0.2838	0.27	350	0.2710
0.2744	0.31	400	0.2634
0.2657	0.35	450	0.2577
0.2692	0.39	500	0.2513
0.263	0.43	550	0.2475
0.2664	0.47	600	0.2451
0.2535	0.51	650	0.2421
0.2594	0.55	700	0.2396
0.234	0.59	750	0.2379
0.2383	0.63	800	0.2361
0.2419	0.67	850	0.2350
0.2448	0.71	900	0.2337
0.241	0.74	950	0.2332
0.219	0.78	1000	0.2330

Framework versions

• PEFT 0.7.2.dev0
• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0