mhqa-cross-encoder-reranker-v2

This model is a fine-tuned version of xlm-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0068

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.0141	0.0217	250	0.0117
0.017	0.0435	500	0.0103
0.0137	0.0652	750	0.0104
0.0156	0.0869	1000	0.0100
0.0137	0.1086	1250	0.0099
0.0128	0.1304	1500	0.0097
0.0131	0.1521	1750	0.0091
0.0103	0.1738	2000	0.0099
0.0096	0.1955	2250	0.0093
0.011	0.2173	2500	0.0092
0.0094	0.2390	2750	0.0089
0.0101	0.2607	3000	0.0092
0.0143	0.2824	3250	0.0088
0.0118	0.3042	3500	0.0093
0.013	0.3259	3750	0.0086
0.0099	0.3476	4000	0.0086
0.0112	0.3693	4250	0.0093
0.0105	0.3911	4500	0.0088
0.0079	0.4128	4750	0.0090
0.0094	0.4345	5000	0.0083
0.0105	0.4562	5250	0.0083
0.0087	0.4780	5500	0.0096
0.0077	0.4997	5750	0.0085
0.0106	0.5214	6000	0.0089
0.0143	0.5431	6250	0.0093
0.0112	0.5649	6500	0.0087
0.0087	0.5866	6750	0.0084
0.0071	0.6083	7000	0.0092
0.0092	0.6301	7250	0.0081
0.0112	0.6518	7500	0.0086
0.0099	0.6735	7750	0.0080
0.0082	0.6952	8000	0.0080
0.0092	0.7170	8250	0.0084
0.0101	0.7387	8500	0.0083
0.0057	0.7604	8750	0.0089
0.0098	0.7821	9000	0.0082
0.0105	0.8039	9250	0.0083
0.0093	0.8256	9500	0.0081
0.0088	0.8473	9750	0.0086
0.0072	0.8690	10000	0.0083
0.008	0.8908	10250	0.0077
0.007	0.9125	10500	0.0081
0.0073	0.9342	10750	0.0077
0.0124	0.9559	11000	0.0074
0.0071	0.9777	11250	0.0077
0.0082	0.9994	11500	0.0079
0.0091	1.0211	11750	0.0077
0.0075	1.0428	12000	0.0080
0.0086	1.0646	12250	0.0075
0.0079	1.0863	12500	0.0085
0.0074	1.1080	12750	0.0074
0.01	1.1297	13000	0.0093
0.0063	1.1515	13250	0.0074
0.0059	1.1732	13500	0.0089
0.0061	1.1949	13750	0.0075
0.0058	1.2167	14000	0.0075
0.0086	1.2384	14250	0.0081
0.0103	1.2601	14500	0.0076
0.0067	1.2818	14750	0.0080
0.0076	1.3036	15000	0.0089
0.0083	1.3253	15250	0.0073
0.0079	1.3470	15500	0.0073
0.0088	1.3687	15750	0.0079
0.0049	1.3905	16000	0.0079
0.0089	1.4122	16250	0.0074
0.0075	1.4339	16500	0.0072
0.0074	1.4556	16750	0.0071
0.0074	1.4774	17000	0.0073
0.008	1.4991	17250	0.0076
0.007	1.5208	17500	0.0071
0.0075	1.5425	17750	0.0071
0.0045	1.5643	18000	0.0071
0.0074	1.5860	18250	0.0072
0.0083	1.6077	18500	0.0075
0.0076	1.6294	18750	0.0071
0.005	1.6512	19000	0.0072
0.007	1.6729	19250	0.0074
0.0068	1.6946	19500	0.0073
0.0066	1.7163	19750	0.0074
0.0069	1.7381	20000	0.0072
0.0075	1.7598	20250	0.0073
0.0074	1.7815	20500	0.0071
0.0071	1.8033	20750	0.0073
0.0087	1.8250	21000	0.0073
0.0053	1.8467	21250	0.0068
0.0054	1.8684	21500	0.0074
0.0058	1.8902	21750	0.0074
0.0077	1.9119	22000	0.0070
0.0061	1.9336	22250	0.0069
0.0078	1.9553	22500	0.0071
0.0055	1.9771	22750	0.0070
0.0048	1.9988	23000	0.0070

Framework versions

• Transformers 4.46.0
• Pytorch 2.12.0+cu130
• Datasets 5.0.0
• Tokenizers 0.20.3