google-siglip-base-patch16-224-batch64-lr5e-05-standford-dogs

This model is a fine-tuned version of google/siglip-base-patch16-224 on the stanford-dogs dataset. It achieves the following results on the evaluation set:

• Loss: 0.5612
• Accuracy: 0.8365
• F1: 0.8329
• Precision: 0.8377
• Recall: 0.8331

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
4.822	0.1550	10	4.2549	0.0782	0.0493	0.0987	0.0726
4.236	0.3101	20	3.5279	0.1907	0.1507	0.2201	0.1830
3.5066	0.4651	30	2.5316	0.3319	0.2941	0.4180	0.3205
2.8064	0.6202	40	2.1243	0.4361	0.4090	0.5324	0.4282
2.441	0.7752	50	1.5798	0.5510	0.5250	0.6242	0.5438
2.0985	0.9302	60	1.4242	0.5843	0.5577	0.6400	0.5768
1.8689	1.0853	70	1.1481	0.6625	0.6456	0.7143	0.6565
1.6588	1.2403	80	1.1937	0.6465	0.6361	0.7062	0.6439
1.5807	1.3953	90	0.9818	0.7058	0.6890	0.7438	0.6981
1.4851	1.5504	100	1.0181	0.7000	0.6839	0.7373	0.6959
1.5033	1.7054	110	1.0169	0.6914	0.6845	0.7490	0.6883
1.3022	1.8605	120	0.9087	0.7276	0.7170	0.7643	0.7222
1.3106	2.0155	130	0.8385	0.7432	0.7352	0.7667	0.7363
1.1721	2.1705	140	0.8957	0.7128	0.7026	0.7592	0.7075
1.131	2.3256	150	0.8730	0.7259	0.7149	0.7687	0.7196
1.1223	2.4806	160	0.8132	0.7546	0.7457	0.7855	0.7482
1.0688	2.6357	170	0.7485	0.7704	0.7601	0.7863	0.7631
1.0686	2.7907	180	0.7559	0.7651	0.7587	0.7920	0.7609
0.9733	2.9457	190	0.7779	0.7553	0.7458	0.7797	0.7521
0.9287	3.1008	200	0.7048	0.7818	0.7756	0.7981	0.7756
0.8746	3.2558	210	0.6848	0.7867	0.7774	0.8034	0.7822
0.7982	3.4109	220	0.6930	0.7884	0.7796	0.8025	0.7846
0.823	3.5659	230	0.7041	0.7804	0.7717	0.7975	0.7752
0.8713	3.7209	240	0.7418	0.7755	0.7646	0.8053	0.7711
0.8651	3.8760	250	0.6847	0.7828	0.7773	0.8048	0.7782
0.784	4.0310	260	0.6662	0.7923	0.7841	0.8097	0.7860
0.6894	4.1860	270	0.6980	0.7843	0.7781	0.8024	0.7779
0.7727	4.3411	280	0.6629	0.7833	0.7804	0.8030	0.7798
0.6978	4.4961	290	0.6820	0.7845	0.7800	0.8011	0.7820
0.7032	4.6512	300	0.6148	0.8032	0.7969	0.8094	0.7985
0.6978	4.8062	310	0.6457	0.7940	0.7872	0.8085	0.7892
0.66	4.9612	320	0.6242	0.8088	0.8033	0.8246	0.8058
0.5706	5.1163	330	0.6404	0.7966	0.7905	0.8097	0.7928
0.5456	5.2713	340	0.7147	0.7872	0.7767	0.8060	0.7819
0.5869	5.4264	350	0.6267	0.8066	0.8016	0.8188	0.8025
0.6022	5.5814	360	0.6197	0.8061	0.8028	0.8209	0.8027
0.5676	5.7364	370	0.6061	0.8059	0.8005	0.8140	0.8024
0.5456	5.8915	380	0.6018	0.8069	0.8006	0.8254	0.8033
0.56	6.0465	390	0.6126	0.8090	0.8037	0.8206	0.8045
0.4582	6.2016	400	0.6122	0.8115	0.8062	0.8196	0.8061
0.4594	6.3566	410	0.6058	0.8122	0.8081	0.8235	0.8082
0.4868	6.5116	420	0.5890	0.8195	0.8131	0.8300	0.8141
0.4841	6.6667	430	0.5909	0.8175	0.8119	0.8250	0.8133
0.4537	6.8217	440	0.5889	0.8195	0.8153	0.8261	0.8164
0.4807	6.9767	450	0.6105	0.8144	0.8104	0.8300	0.8106
0.4051	7.1318	460	0.5917	0.8171	0.8103	0.8217	0.8131
0.3727	7.2868	470	0.6037	0.8166	0.8116	0.8262	0.8125
0.4034	7.4419	480	0.6407	0.8032	0.8003	0.8146	0.8015
0.3684	7.5969	490	0.6205	0.8061	0.7997	0.8176	0.8008
0.416	7.7519	500	0.5855	0.8258	0.8207	0.8364	0.8211
0.3947	7.9070	510	0.5802	0.8214	0.8179	0.8283	0.8179
0.3731	8.0620	520	0.5870	0.8239	0.8191	0.8324	0.8188
0.3203	8.2171	530	0.5783	0.8265	0.8211	0.8302	0.8216
0.337	8.3721	540	0.5836	0.8200	0.8162	0.8247	0.8166
0.3396	8.5271	550	0.5992	0.8156	0.8121	0.8253	0.8115
0.3355	8.6822	560	0.5755	0.8229	0.8182	0.8281	0.8187
0.3273	8.8372	570	0.5819	0.8246	0.8194	0.8268	0.8208
0.3181	8.9922	580	0.5840	0.8205	0.8174	0.8279	0.8168
0.2855	9.1473	590	0.5997	0.8144	0.8098	0.8213	0.8103
0.254	9.3023	600	0.5863	0.8183	0.8132	0.8251	0.8133
0.2781	9.4574	610	0.5779	0.8224	0.8169	0.8275	0.8195
0.2691	9.6124	620	0.5816	0.8219	0.8177	0.8257	0.8186
0.3018	9.7674	630	0.5814	0.8297	0.8250	0.8370	0.8253
0.2615	9.9225	640	0.5761	0.8299	0.8261	0.8377	0.8262
0.2707	10.0775	650	0.5640	0.8326	0.8283	0.8385	0.8284
0.2482	10.2326	660	0.5685	0.8246	0.8206	0.8284	0.8218
0.2493	10.3876	670	0.5717	0.8241	0.8208	0.8311	0.8199
0.2167	10.5426	680	0.5741	0.8246	0.8204	0.8273	0.8204
0.2628	10.6977	690	0.5791	0.8248	0.8205	0.8281	0.8216
0.2316	10.8527	700	0.5770	0.8321	0.8272	0.8348	0.8284
0.2326	11.0078	710	0.5755	0.8280	0.8249	0.8348	0.8249
0.2001	11.1628	720	0.5783	0.8336	0.8299	0.8354	0.8310
0.1759	11.3178	730	0.5804	0.8345	0.8302	0.8367	0.8311
0.202	11.4729	740	0.5820	0.8316	0.8278	0.8353	0.8280
0.2191	11.6279	750	0.5724	0.8324	0.8279	0.8341	0.8287
0.1955	11.7829	760	0.5957	0.8226	0.8181	0.8268	0.8198
0.1972	11.9380	770	0.5722	0.8294	0.8254	0.8318	0.8263
0.1848	12.0930	780	0.5731	0.8311	0.8269	0.8339	0.8281
0.1613	12.2481	790	0.5682	0.8382	0.8344	0.8397	0.8356
0.1665	12.4031	800	0.5565	0.8350	0.8325	0.8365	0.8325
0.1739	12.5581	810	0.5738	0.8360	0.8328	0.8395	0.8326
0.1744	12.7132	820	0.5628	0.8360	0.8327	0.8387	0.8328
0.1737	12.8682	830	0.5712	0.8355	0.8320	0.8395	0.8324
0.1635	13.0233	840	0.5745	0.8309	0.8256	0.8328	0.8269
0.1689	13.1783	850	0.5781	0.8326	0.8288	0.8358	0.8294
0.1611	13.3333	860	0.5740	0.8328	0.8280	0.8349	0.8289
0.1624	13.4884	870	0.5656	0.8324	0.8279	0.8328	0.8287
0.1635	13.6434	880	0.5618	0.8319	0.8276	0.8328	0.8280
0.1395	13.7984	890	0.5648	0.8350	0.8311	0.8368	0.8312
0.1489	13.9535	900	0.5666	0.8341	0.8304	0.8370	0.8304
0.1174	14.1085	910	0.5700	0.8358	0.8321	0.8400	0.8320
0.1274	14.2636	920	0.5720	0.8331	0.8295	0.8366	0.8295
0.134	14.4186	930	0.5657	0.8353	0.8311	0.8369	0.8317
0.1327	14.5736	940	0.5662	0.8343	0.8308	0.8367	0.8307
0.1165	14.7287	950	0.5654	0.8341	0.8301	0.8355	0.8303
0.1277	14.8837	960	0.5661	0.8345	0.8308	0.8360	0.8310
0.1221	15.0388	970	0.5615	0.8370	0.8335	0.8388	0.8335
0.1194	15.1938	980	0.5632	0.8353	0.8318	0.8369	0.8319
0.1126	15.3488	990	0.5616	0.8362	0.8326	0.8376	0.8327
0.1256	15.5039	1000	0.5612	0.8365	0.8329	0.8377	0.8331

Framework versions

• Transformers 4.40.2
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1