ast-arabic

This model is a fine-tuned version of MIT/ast-finetuned-speech-commands-v2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6961
• Accuracy: 0.7871
• Precision: 0.8186
• Recall: 0.7871
• F1: 0.7848
• Binary: 0.8501

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Binary
No log	0.19	50	4.5626	0.0027	0.0001	0.0027	0.0003	0.0860
No log	0.38	100	4.4915	0.0054	0.0037	0.0054	0.0034	0.0954
No log	0.58	150	4.3465	0.0216	0.0105	0.0216	0.0123	0.1488
No log	0.77	200	4.1317	0.0647	0.0355	0.0647	0.0372	0.2356
No log	0.96	250	3.8709	0.1267	0.0963	0.1267	0.0880	0.3245
No log	1.15	300	3.4181	0.2318	0.2405	0.2318	0.1927	0.4442
No log	1.34	350	2.8697	0.3612	0.3714	0.3612	0.3150	0.5426
No log	1.53	400	2.3609	0.4663	0.4425	0.4663	0.4157	0.6248
No log	1.73	450	1.9516	0.5606	0.5573	0.5606	0.5246	0.6903
3.5112	1.92	500	1.6670	0.6092	0.6428	0.6092	0.5849	0.7240
3.5112	2.11	550	1.4376	0.6739	0.6988	0.6739	0.6544	0.7709
3.5112	2.3	600	1.3049	0.6658	0.7019	0.6658	0.6495	0.7636
3.5112	2.49	650	1.1820	0.6846	0.7214	0.6846	0.6731	0.7784
3.5112	2.68	700	1.0847	0.7035	0.7220	0.7035	0.6913	0.7916
3.5112	2.88	750	1.0370	0.7116	0.7415	0.7116	0.7004	0.7973
3.5112	3.07	800	0.9833	0.7062	0.7132	0.7062	0.6903	0.7935
3.5112	3.26	850	0.9453	0.7116	0.7384	0.7116	0.6986	0.7973
3.5112	3.45	900	0.9140	0.7412	0.7724	0.7412	0.7313	0.8181
3.5112	3.64	950	0.8602	0.7493	0.7719	0.7493	0.7403	0.8237
1.0914	3.84	1000	0.8340	0.7520	0.7857	0.7520	0.7469	0.8256
1.0914	4.03	1050	0.8317	0.7628	0.8014	0.7628	0.7602	0.8332
1.0914	4.22	1100	0.7983	0.7628	0.7863	0.7628	0.7584	0.8332
1.0914	4.41	1150	0.8015	0.7601	0.7852	0.7601	0.7529	0.8313
1.0914	4.6	1200	0.7584	0.7709	0.7959	0.7709	0.7685	0.8388
1.0914	4.79	1250	0.7518	0.7763	0.7957	0.7763	0.7699	0.8426
1.0914	4.99	1300	0.7484	0.7520	0.7698	0.7520	0.7449	0.8256
1.0914	5.18	1350	0.7518	0.7466	0.7711	0.7466	0.7397	0.8218
1.0914	5.37	1400	0.7379	0.7682	0.7961	0.7682	0.7654	0.8369
1.0914	5.56	1450	0.7356	0.7601	0.7770	0.7601	0.7533	0.8313
0.6112	5.75	1500	0.7351	0.7763	0.8018	0.7763	0.7720	0.8426
0.6112	5.94	1550	0.7230	0.7655	0.7886	0.7655	0.7631	0.8350
0.6112	6.14	1600	0.7222	0.7709	0.7977	0.7709	0.7656	0.8388
0.6112	6.33	1650	0.7054	0.7790	0.8022	0.7790	0.7744	0.8445
0.6112	6.52	1700	0.7286	0.7736	0.8016	0.7736	0.7693	0.8407
0.6112	6.71	1750	0.6991	0.7925	0.8141	0.7925	0.7907	0.8539
0.6112	6.9	1800	0.7096	0.7655	0.7907	0.7655	0.7633	0.8350
0.6112	7.09	1850	0.7010	0.7844	0.8062	0.7844	0.7821	0.8482
0.6112	7.29	1900	0.7026	0.7574	0.7813	0.7574	0.7514	0.8294
0.6112	7.48	1950	0.6973	0.7817	0.7992	0.7817	0.7785	0.8464
0.4147	7.67	2000	0.7035	0.7736	0.7911	0.7736	0.7707	0.8407
0.4147	7.86	2050	0.6894	0.7844	0.8089	0.7844	0.7827	0.8482
0.4147	8.05	2100	0.7046	0.7763	0.8009	0.7763	0.7731	0.8426
0.4147	8.25	2150	0.6994	0.7817	0.8028	0.7817	0.7786	0.8464
0.4147	8.44	2200	0.6890	0.7763	0.7994	0.7763	0.7723	0.8426
0.4147	8.63	2250	0.6988	0.7763	0.7954	0.7763	0.7722	0.8426
0.4147	8.82	2300	0.7021	0.7763	0.7964	0.7763	0.7731	0.8426
0.4147	9.01	2350	0.6974	0.7817	0.8048	0.7817	0.7784	0.8464
0.4147	9.2	2400	0.6861	0.7817	0.8041	0.7817	0.7807	0.8464
0.4147	9.4	2450	0.6930	0.7763	0.7994	0.7763	0.7734	0.8426
0.3165	9.59	2500	0.6954	0.7763	0.8001	0.7763	0.7741	0.8426
0.3165	9.78	2550	0.6953	0.7817	0.8045	0.7817	0.7794	0.8464
0.3165	9.97	2600	0.6940	0.7817	0.8053	0.7817	0.7794	0.8464

Framework versions

• Transformers 4.38.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.15.1