Llama-3.2-1B-Instruct_v2

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2356

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 30
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss
0.4565	0.0333	100	0.4058
0.3313	0.0667	200	0.3428
0.3139	0.1	300	0.3192
0.2903	0.1333	400	0.3034
0.2639	0.1667	500	0.2944
0.2688	0.2	600	0.2869
0.3097	0.2333	700	0.2791
0.2462	0.2667	800	0.2735
0.3257	0.3	900	0.2684
0.2738	0.3333	1000	0.2638
0.2572	0.3667	1100	0.2598
0.234	0.4	1200	0.2566
0.2233	0.4333	1300	0.2537
0.2996	0.4667	1400	0.2515
0.2178	0.5	1500	0.2490
0.2251	0.5333	1600	0.2470
0.262	0.5667	1700	0.2450
0.2683	0.6	1800	0.2430
0.1966	0.6333	1900	0.2416
0.2451	0.6667	2000	0.2403
0.2247	0.7	2100	0.2393
0.1865	0.7333	2200	0.2384
0.2837	0.7667	2300	0.2378
0.2312	0.8	2400	0.2371
0.239	0.8333	2500	0.2365
0.2064	0.8667	2600	0.2362
0.208	0.9	2700	0.2358
0.2588	0.9333	2800	0.2356
0.2029	0.9667	2900	0.2356
0.2404	1.0	3000	0.2356

Framework versions

• PEFT 0.13.2
• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1