xlm-roberta-xl-lora4

This model is a fine-tuned version of facebook/xlm-roberta-xl on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2790
• Precision: 0.9287
• Recall: 0.9301
• F1: 0.9294
• Accuracy: 0.9392

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 64
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 63
• num_epochs: 50
• label_smoothing_factor: 0.15

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
3.6153	1.0	63	2.8304	0.5449	0.6685	0.6004	0.6459
2.4657	2.0	126	2.1379	0.7373	0.8004	0.7676	0.8065
1.9364	3.0	189	1.7624	0.8330	0.8617	0.8472	0.8740
1.6441	4.0	252	1.5723	0.8663	0.8816	0.8739	0.8949
1.494	5.0	315	1.4820	0.8779	0.8884	0.8831	0.9020
1.3987	6.0	378	1.4190	0.8961	0.9012	0.8987	0.9135
1.3388	7.0	441	1.3814	0.9023	0.9073	0.9048	0.9187
1.2947	8.0	504	1.3609	0.8976	0.9082	0.9029	0.9200
1.2585	9.0	567	1.3415	0.8965	0.9113	0.9038	0.9203
1.2317	10.0	630	1.3246	0.9095	0.9095	0.9095	0.9246
1.2081	11.0	693	1.3111	0.9095	0.9143	0.9119	0.9268
1.1869	12.0	756	1.3005	0.9161	0.9194	0.9177	0.9305
1.1711	13.0	819	1.3085	0.9069	0.9169	0.9119	0.9265
1.1557	14.0	882	1.2989	0.9191	0.9204	0.9198	0.9309
1.1486	15.0	945	1.2962	0.9166	0.9185	0.9176	0.9295
1.1392	16.0	1008	1.2796	0.9202	0.9228	0.9215	0.9348
1.127	17.0	1071	1.2830	0.9200	0.9229	0.9214	0.9341
1.1224	18.0	1134	1.2814	0.9184	0.9248	0.9216	0.9336
1.1146	19.0	1197	1.2775	0.9206	0.9260	0.9233	0.9356
1.1081	20.0	1260	1.2798	0.9251	0.9263	0.9257	0.9358
1.1006	21.0	1323	1.2756	0.9220	0.9257	0.9238	0.9364
1.0972	22.0	1386	1.2755	0.9176	0.9258	0.9217	0.9357
1.0926	23.0	1449	1.2795	0.9217	0.9267	0.9242	0.9366
1.0898	24.0	1512	1.2830	0.9213	0.9260	0.9236	0.9348
1.0847	25.0	1575	1.2749	0.9234	0.9275	0.9255	0.9377
1.0818	26.0	1638	1.2806	0.9245	0.9270	0.9257	0.9368
1.0796	27.0	1701	1.2760	0.9243	0.9283	0.9263	0.9372
1.0753	28.0	1764	1.2776	0.9220	0.9264	0.9242	0.9364
1.072	29.0	1827	1.2755	0.9265	0.9288	0.9276	0.9388
1.0686	30.0	1890	1.2752	0.9240	0.9246	0.9243	0.9365
1.0676	31.0	1953	1.2755	0.9271	0.9293	0.9282	0.9386
1.0663	32.0	2016	1.2771	0.9261	0.9282	0.9272	0.9383
1.0646	33.0	2079	1.2774	0.9235	0.9283	0.9259	0.9370
1.0641	34.0	2142	1.2710	0.9274	0.9313	0.9294	0.9398
1.0648	35.0	2205	1.2759	0.9259	0.9284	0.9271	0.9387
1.0623	36.0	2268	1.2741	0.9260	0.9294	0.9277	0.9383
1.06	37.0	2331	1.2747	0.9243	0.9293	0.9268	0.9377
1.0592	38.0	2394	1.2757	0.9262	0.9293	0.9278	0.9389
1.0581	39.0	2457	1.2794	0.9251	0.9294	0.9273	0.9379
1.0574	40.0	2520	1.2765	0.9295	0.9298	0.9296	0.9400
1.0569	41.0	2583	1.2798	0.9253	0.9281	0.9267	0.9381
1.0557	42.0	2646	1.2813	0.9282	0.9294	0.9288	0.9391
1.0562	43.0	2709	1.2792	0.9253	0.9261	0.9257	0.9366
1.056	44.0	2772	1.2797	0.9266	0.9293	0.9280	0.9386
1.0545	45.0	2835	1.2800	0.9265	0.9284	0.9274	0.9382
1.0546	46.0	2898	1.2788	0.9284	0.9299	0.9292	0.9394
1.0544	47.0	2961	1.2794	0.9280	0.9292	0.9286	0.9386
1.0539	48.0	3024	1.2785	0.9285	0.9299	0.9292	0.9393
1.054	49.0	3087	1.2791	0.9284	0.9294	0.9289	0.9390
1.0538	50.0	3150	1.2790	0.9287	0.9301	0.9294	0.9392

Framework versions

• Transformers 4.31.0
• Pytorch 2.1.0
• Datasets 2.14.5
• Tokenizers 0.13.3