TiramisuQiao/Llama-3.1-8B-Ins-SFT-nnetnav-wa

output

This model is a fine-tuned version of Llama-3.1-8B-Instruct on the webarena_NNetNav_Bof3 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7705

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• total_eval_batch_size: 8
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
0.9481	0.0393	50	0.8577
0.9427	0.0785	100	0.8540
0.8686	0.1178	150	0.8634
0.9135	0.1570	200	0.8639
0.8547	0.1963	250	0.8516
0.9755	0.2356	300	0.8471
0.9139	0.2748	350	0.8421
0.8972	0.3141	400	0.8355
0.8496	0.3534	450	0.8296
0.9462	0.3926	500	0.8225
0.8491	0.4319	550	0.8206
0.9046	0.4711	600	0.8149
0.8733	0.5104	650	0.8085
0.8544	0.5497	700	0.8022
0.8321	0.5889	750	0.7972
0.8463	0.6282	800	0.7934
0.8954	0.6675	850	0.7889
0.7878	0.7067	900	0.7845
0.8202	0.7460	950	0.7816
0.8099	0.7852	1000	0.7781
0.8112	0.8245	1050	0.7752
0.8273	0.8638	1100	0.7729
0.8141	0.9030	1150	0.7715
0.8065	0.9423	1200	0.7707
0.8345	0.9815	1250	0.7704

Framework versions

• Transformers 4.55.0
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1