End of training

all_results.json

@@ -0,0 +1,14 @@
+{
+    "epoch": 16.0,
+    "eval_loss": 0.46361204981803894,
+    "eval_matthews_correlation": 0.5057060886900621,
+    "eval_runtime": 1.1268,
+    "eval_samples": 1043,
+    "eval_samples_per_second": 925.6,
+    "eval_steps_per_second": 58.571,
+    "train_loss": 0.18462801172133925,
+    "train_runtime": 361.7714,
+    "train_samples": 8551,
+    "train_samples_per_second": 378.184,
+    "train_steps_per_second": 2.963
+}

eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 16.0,
+    "eval_loss": 0.46361204981803894,
+    "eval_matthews_correlation": 0.5057060886900621,
+    "eval_runtime": 1.1268,
+    "eval_samples": 1043,
+    "eval_samples_per_second": 925.6,
+    "eval_steps_per_second": 58.571
+}

predict_results_cola.txt

@@ -0,0 +1,1064 @@
+index	prediction
+0	acceptable
+1	acceptable
+2	acceptable
+3	acceptable
+4	unacceptable
+5	acceptable
+6	acceptable
+7	unacceptable
+8	unacceptable
+9	acceptable
+10	acceptable
+11	unacceptable
+12	unacceptable
+13	acceptable
+14	acceptable
+15	acceptable
+16	unacceptable
+17	acceptable
+18	acceptable
+19	unacceptable
+20	acceptable
+21	acceptable
+22	acceptable
+23	acceptable
+24	acceptable
+25	acceptable
+26	acceptable
+27	unacceptable
+28	unacceptable
+29	acceptable
+30	unacceptable
+31	unacceptable
+32	acceptable
+33	acceptable
+34	unacceptable
+35	acceptable
+36	acceptable
+37	unacceptable
+38	unacceptable
+39	unacceptable
+40	unacceptable
+41	unacceptable
+42	unacceptable
+43	acceptable
+44	unacceptable
+45	acceptable
+46	unacceptable
+47	unacceptable
+48	acceptable
+49	acceptable
+50	unacceptable
+51	unacceptable
+52	acceptable
+53	acceptable
+54	acceptable
+55	acceptable
+56	acceptable
+57	unacceptable
+58	acceptable
+59	acceptable
+60	unacceptable
+61	acceptable
+62	acceptable
+63	acceptable
+64	acceptable
+65	acceptable
+66	acceptable
+67	acceptable
+68	acceptable
+69	acceptable
+70	acceptable
+71	unacceptable
+72	unacceptable
+73	unacceptable
+74	acceptable
+75	acceptable
+76	acceptable
+77	acceptable
+78	acceptable
+79	acceptable
+80	acceptable
+81	unacceptable
+82	unacceptable
+83	acceptable
+84	unacceptable
+85	acceptable
+86	acceptable
+87	unacceptable
+88	acceptable
+89	acceptable
+90	unacceptable
+91	acceptable
+92	acceptable
+93	acceptable
+94	acceptable
+95	acceptable
+96	acceptable
+97	acceptable
+98	acceptable
+99	unacceptable
+100	unacceptable
+101	unacceptable
+102	acceptable
+103	acceptable
+104	acceptable
+105	acceptable
+106	acceptable
+107	acceptable
+108	acceptable
+109	unacceptable
+110	acceptable
+111	acceptable
+112	acceptable
+113	acceptable
+114	acceptable
+115	unacceptable
+116	acceptable
+117	unacceptable
+118	unacceptable
+119	acceptable
+120	unacceptable
+121	acceptable
+122	unacceptable
+123	unacceptable
+124	acceptable
+125	acceptable
+126	unacceptable
+127	acceptable
+128	acceptable
+129	acceptable
+130	unacceptable
+131	acceptable
+132	acceptable
+133	acceptable
+134	acceptable
+135	acceptable
+136	acceptable
+137	acceptable
+138	unacceptable
+139	unacceptable
+140	unacceptable
+141	unacceptable
+142	acceptable
+143	acceptable
+144	acceptable
+145	acceptable
+146	acceptable
+147	unacceptable
+148	acceptable
+149	acceptable
+150	acceptable
+151	acceptable
+152	unacceptable
+153	acceptable
+154	acceptable
+155	unacceptable
+156	unacceptable
+157	unacceptable
+158	acceptable
+159	acceptable
+160	acceptable
+161	acceptable
+162	acceptable
+163	unacceptable
+164	acceptable
+165	unacceptable
+166	unacceptable
+167	unacceptable
+168	acceptable
+169	acceptable
+170	acceptable
+171	acceptable
+172	acceptable
+173	unacceptable
+174	acceptable
+175	acceptable
+176	acceptable
+177	acceptable
+178	acceptable
+179	acceptable
+180	acceptable
+181	acceptable
+182	unacceptable
+183	acceptable
+184	unacceptable
+185	acceptable
+186	acceptable
+187	acceptable
+188	unacceptable
+189	acceptable
+190	acceptable
+191	acceptable
+192	acceptable
+193	acceptable
+194	unacceptable
+195	acceptable
+196	unacceptable
+197	unacceptable
+198	acceptable
+199	acceptable
+200	acceptable
+201	unacceptable
+202	unacceptable
+203	acceptable
+204	acceptable
+205	acceptable
+206	acceptable
+207	acceptable
+208	acceptable
+209	acceptable
+210	acceptable
+211	acceptable
+212	acceptable
+213	unacceptable
+214	acceptable
+215	acceptable
+216	acceptable
+217	acceptable
+218	acceptable
+219	acceptable
+220	acceptable
+221	unacceptable
+222	acceptable
+223	acceptable
+224	unacceptable
+225	acceptable
+226	unacceptable
+227	acceptable
+228	unacceptable
+229	acceptable
+230	acceptable
+231	acceptable
+232	acceptable
+233	acceptable
+234	acceptable
+235	acceptable
+236	unacceptable
+237	acceptable
+238	acceptable
+239	unacceptable
+240	acceptable
+241	acceptable
+242	acceptable
+243	acceptable
+244	acceptable
+245	acceptable
+246	acceptable
+247	acceptable
+248	acceptable
+249	acceptable
+250	acceptable
+251	acceptable
+252	acceptable
+253	acceptable
+254	acceptable
+255	acceptable
+256	acceptable
+257	acceptable
+258	unacceptable
+259	acceptable
+260	unacceptable
+261	acceptable
+262	acceptable
+263	acceptable
+264	acceptable
+265	acceptable
+266	acceptable
+267	acceptable
+268	acceptable
+269	acceptable
+270	acceptable
+271	acceptable
+272	unacceptable
+273	acceptable
+274	acceptable
+275	acceptable
+276	acceptable
+277	unacceptable
+278	acceptable
+279	unacceptable
+280	acceptable
+281	acceptable
+282	acceptable
+283	acceptable
+284	acceptable
+285	unacceptable
+286	unacceptable
+287	acceptable
+288	acceptable
+289	acceptable
+290	acceptable
+291	acceptable
+292	acceptable
+293	unacceptable
+294	unacceptable
+295	acceptable
+296	acceptable
+297	acceptable
+298	acceptable
+299	acceptable
+300	acceptable
+301	unacceptable
+302	acceptable
+303	acceptable
+304	acceptable
+305	acceptable
+306	unacceptable
+307	acceptable
+308	acceptable
+309	acceptable
+310	acceptable
+311	acceptable
+312	acceptable
+313	acceptable
+314	unacceptable
+315	unacceptable
+316	acceptable
+317	unacceptable
+318	acceptable
+319	unacceptable
+320	unacceptable
+321	acceptable
+322	unacceptable
+323	acceptable
+324	acceptable
+325	acceptable
+326	acceptable
+327	acceptable
+328	acceptable
+329	acceptable
+330	acceptable
+331	acceptable
+332	acceptable
+333	acceptable
+334	acceptable
+335	acceptable
+336	acceptable
+337	acceptable
+338	acceptable
+339	acceptable
+340	acceptable
+341	acceptable
+342	unacceptable
+343	unacceptable
+344	unacceptable
+345	acceptable
+346	acceptable
+347	acceptable
+348	unacceptable
+349	acceptable
+350	acceptable
+351	acceptable
+352	acceptable
+353	acceptable
+354	acceptable
+355	acceptable
+356	unacceptable
+357	acceptable
+358	acceptable
+359	acceptable
+360	acceptable
+361	unacceptable
+362	acceptable
+363	acceptable
+364	acceptable
+365	unacceptable
+366	acceptable
+367	acceptable
+368	unacceptable
+369	acceptable
+370	acceptable
+371	acceptable
+372	acceptable
+373	acceptable
+374	acceptable
+375	acceptable
+376	acceptable
+377	acceptable
+378	acceptable
+379	acceptable
+380	acceptable
+381	acceptable
+382	acceptable
+383	acceptable
+384	acceptable
+385	acceptable
+386	acceptable
+387	acceptable
+388	acceptable
+389	unacceptable
+390	acceptable
+391	unacceptable
+392	unacceptable
+393	acceptable
+394	acceptable
+395	acceptable
+396	acceptable
+397	acceptable
+398	acceptable
+399	acceptable
+400	acceptable
+401	acceptable
+402	unacceptable
+403	acceptable
+404	acceptable
+405	acceptable
+406	acceptable
+407	acceptable
+408	acceptable
+409	acceptable
+410	acceptable
+411	acceptable
+412	acceptable
+413	acceptable
+414	acceptable
+415	acceptable
+416	acceptable
+417	unacceptable
+418	unacceptable
+419	unacceptable
+420	acceptable
+421	acceptable
+422	acceptable
+423	unacceptable
+424	acceptable
+425	acceptable
+426	acceptable
+427	acceptable
+428	acceptable
+429	unacceptable
+430	acceptable
+431	unacceptable
+432	acceptable
+433	acceptable
+434	acceptable
+435	unacceptable
+436	acceptable
+437	acceptable
+438	acceptable
+439	acceptable
+440	acceptable
+441	acceptable
+442	acceptable
+443	acceptable
+444	unacceptable
+445	unacceptable
+446	unacceptable
+447	acceptable
+448	unacceptable
+449	acceptable
+450	acceptable
+451	acceptable
+452	acceptable
+453	acceptable
+454	acceptable
+455	unacceptable
+456	acceptable
+457	acceptable
+458	acceptable
+459	unacceptable
+460	acceptable
+461	acceptable
+462	acceptable
+463	acceptable
+464	acceptable
+465	acceptable
+466	acceptable
+467	acceptable
+468	acceptable
+469	acceptable
+470	acceptable
+471	acceptable
+472	acceptable
+473	unacceptable
+474	acceptable
+475	unacceptable
+476	unacceptable
+477	acceptable
+478	acceptable
+479	unacceptable
+480	acceptable
+481	unacceptable
+482	acceptable
+483	unacceptable
+484	acceptable
+485	acceptable
+486	acceptable
+487	acceptable
+488	unacceptable
+489	unacceptable
+490	acceptable
+491	unacceptable
+492	acceptable
+493	acceptable
+494	acceptable
+495	unacceptable
+496	unacceptable
+497	acceptable
+498	acceptable
+499	acceptable
+500	acceptable
+501	acceptable
+502	acceptable
+503	acceptable
+504	unacceptable
+505	acceptable
+506	acceptable
+507	acceptable
+508	unacceptable
+509	acceptable
+510	acceptable
+511	unacceptable
+512	acceptable
+513	acceptable
+514	acceptable
+515	acceptable
+516	acceptable
+517	acceptable
+518	acceptable
+519	acceptable
+520	acceptable
+521	unacceptable
+522	acceptable
+523	acceptable
+524	unacceptable
+525	unacceptable
+526	acceptable
+527	acceptable
+528	acceptable
+529	acceptable
+530	acceptable
+531	unacceptable
+532	acceptable
+533	acceptable
+534	acceptable
+535	acceptable
+536	acceptable
+537	acceptable
+538	acceptable
+539	acceptable
+540	acceptable
+541	acceptable
+542	acceptable
+543	acceptable
+544	unacceptable
+545	acceptable
+546	acceptable
+547	acceptable
+548	acceptable
+549	acceptable
+550	acceptable
+551	acceptable
+552	acceptable
+553	acceptable
+554	acceptable
+555	acceptable
+556	acceptable
+557	acceptable
+558	acceptable
+559	acceptable
+560	acceptable
+561	acceptable
+562	acceptable
+563	acceptable
+564	unacceptable
+565	acceptable
+566	acceptable
+567	acceptable
+568	acceptable
+569	acceptable
+570	unacceptable
+571	unacceptable
+572	acceptable
+573	acceptable
+574	acceptable
+575	unacceptable
+576	acceptable
+577	acceptable
+578	acceptable
+579	acceptable
+580	unacceptable
+581	unacceptable
+582	unacceptable
+583	acceptable
+584	acceptable
+585	acceptable
+586	unacceptable
+587	acceptable
+588	acceptable
+589	acceptable
+590	acceptable
+591	acceptable
+592	unacceptable
+593	acceptable
+594	acceptable
+595	unacceptable
+596	acceptable
+597	acceptable
+598	unacceptable
+599	acceptable
+600	acceptable
+601	unacceptable
+602	acceptable
+603	acceptable
+604	unacceptable
+605	acceptable
+606	acceptable
+607	acceptable
+608	acceptable
+609	acceptable
+610	unacceptable
+611	unacceptable
+612	acceptable
+613	acceptable
+614	acceptable
+615	acceptable
+616	acceptable
+617	acceptable
+618	acceptable
+619	acceptable
+620	acceptable
+621	acceptable
+622	acceptable
+623	acceptable
+624	acceptable
+625	acceptable
+626	acceptable
+627	acceptable
+628	acceptable
+629	acceptable
+630	unacceptable
+631	acceptable
+632	acceptable
+633	unacceptable
+634	acceptable
+635	unacceptable
+636	unacceptable
+637	unacceptable
+638	unacceptable
+639	acceptable
+640	unacceptable
+641	acceptable
+642	unacceptable
+643	acceptable
+644	acceptable
+645	unacceptable
+646	unacceptable
+647	acceptable
+648	unacceptable
+649	unacceptable
+650	acceptable
+651	acceptable
+652	acceptable
+653	acceptable
+654	acceptable
+655	unacceptable
+656	unacceptable
+657	unacceptable
+658	acceptable
+659	unacceptable
+660	acceptable
+661	acceptable
+662	acceptable
+663	acceptable
+664	acceptable
+665	acceptable
+666	acceptable
+667	acceptable
+668	unacceptable
+669	acceptable
+670	acceptable
+671	acceptable
+672	acceptable
+673	acceptable
+674	unacceptable
+675	acceptable
+676	acceptable
+677	unacceptable
+678	acceptable
+679	unacceptable
+680	acceptable
+681	acceptable
+682	acceptable
+683	acceptable
+684	acceptable
+685	acceptable
+686	acceptable
+687	acceptable
+688	unacceptable
+689	unacceptable
+690	unacceptable
+691	acceptable
+692	unacceptable
+693	acceptable
+694	unacceptable
+695	unacceptable
+696	acceptable
+697	acceptable
+698	acceptable
+699	acceptable
+700	acceptable
+701	unacceptable
+702	unacceptable
+703	unacceptable
+704	unacceptable
+705	unacceptable
+706	unacceptable
+707	acceptable
+708	acceptable
+709	acceptable
+710	unacceptable
+711	acceptable
+712	acceptable
+713	acceptable
+714	acceptable
+715	unacceptable
+716	unacceptable
+717	unacceptable
+718	unacceptable
+719	unacceptable
+720	acceptable
+721	unacceptable
+722	unacceptable
+723	unacceptable
+724	unacceptable
+725	unacceptable
+726	acceptable
+727	unacceptable
+728	acceptable
+729	acceptable
+730	acceptable
+731	acceptable
+732	acceptable
+733	acceptable
+734	unacceptable
+735	unacceptable
+736	unacceptable
+737	acceptable
+738	acceptable
+739	acceptable
+740	acceptable
+741	unacceptable
+742	acceptable
+743	acceptable
+744	unacceptable
+745	acceptable
+746	acceptable
+747	unacceptable
+748	acceptable
+749	acceptable
+750	acceptable
+751	acceptable
+752	acceptable
+753	acceptable
+754	acceptable
+755	unacceptable
+756	unacceptable
+757	acceptable
+758	acceptable
+759	acceptable
+760	acceptable
+761	acceptable
+762	acceptable
+763	acceptable
+764	acceptable
+765	acceptable
+766	acceptable
+767	acceptable
+768	acceptable
+769	acceptable
+770	acceptable
+771	acceptable
+772	acceptable
+773	acceptable
+774	acceptable
+775	unacceptable
+776	acceptable
+777	acceptable
+778	acceptable
+779	acceptable
+780	acceptable
+781	acceptable
+782	acceptable
+783	acceptable
+784	acceptable
+785	acceptable
+786	acceptable
+787	acceptable
+788	acceptable
+789	acceptable
+790	acceptable
+791	acceptable
+792	acceptable
+793	acceptable
+794	unacceptable
+795	acceptable
+796	unacceptable
+797	acceptable
+798	acceptable
+799	unacceptable
+800	acceptable
+801	acceptable
+802	acceptable
+803	acceptable
+804	acceptable
+805	acceptable
+806	acceptable
+807	acceptable
+808	unacceptable
+809	acceptable
+810	acceptable
+811	acceptable
+812	acceptable
+813	acceptable
+814	acceptable
+815	unacceptable
+816	acceptable
+817	acceptable
+818	acceptable
+819	acceptable
+820	acceptable
+821	acceptable
+822	unacceptable
+823	unacceptable
+824	acceptable
+825	unacceptable
+826	acceptable
+827	acceptable
+828	acceptable
+829	acceptable
+830	unacceptable
+831	unacceptable
+832	unacceptable
+833	acceptable
+834	acceptable
+835	acceptable
+836	unacceptable
+837	acceptable
+838	acceptable
+839	acceptable
+840	acceptable
+841	unacceptable
+842	acceptable
+843	acceptable
+844	unacceptable
+845	acceptable
+846	unacceptable
+847	acceptable
+848	acceptable
+849	acceptable
+850	acceptable
+851	acceptable
+852	acceptable
+853	acceptable
+854	acceptable
+855	acceptable
+856	acceptable
+857	acceptable
+858	unacceptable
+859	acceptable
+860	acceptable
+861	acceptable
+862	acceptable
+863	acceptable
+864	acceptable
+865	acceptable
+866	acceptable
+867	acceptable
+868	acceptable
+869	acceptable
+870	unacceptable
+871	acceptable
+872	unacceptable
+873	unacceptable
+874	acceptable
+875	acceptable
+876	acceptable
+877	acceptable
+878	acceptable
+879	acceptable
+880	acceptable
+881	acceptable
+882	acceptable
+883	acceptable
+884	acceptable
+885	acceptable
+886	unacceptable
+887	unacceptable
+888	unacceptable
+889	acceptable
+890	unacceptable
+891	unacceptable
+892	unacceptable
+893	acceptable
+894	acceptable
+895	acceptable
+896	acceptable
+897	unacceptable
+898	acceptable
+899	acceptable
+900	acceptable
+901	acceptable
+902	acceptable
+903	acceptable
+904	unacceptable
+905	acceptable
+906	acceptable
+907	acceptable
+908	unacceptable
+909	acceptable
+910	acceptable
+911	acceptable
+912	acceptable
+913	acceptable
+914	acceptable
+915	acceptable
+916	unacceptable
+917	acceptable
+918	acceptable
+919	acceptable
+920	acceptable
+921	acceptable
+922	acceptable
+923	acceptable
+924	acceptable
+925	acceptable
+926	acceptable
+927	acceptable
+928	acceptable
+929	unacceptable
+930	unacceptable
+931	unacceptable
+932	acceptable
+933	unacceptable
+934	acceptable
+935	acceptable
+936	acceptable
+937	unacceptable
+938	acceptable
+939	unacceptable
+940	acceptable
+941	acceptable
+942	acceptable
+943	acceptable
+944	acceptable
+945	acceptable
+946	acceptable
+947	acceptable
+948	unacceptable
+949	unacceptable
+950	acceptable
+951	acceptable
+952	unacceptable
+953	acceptable
+954	acceptable
+955	unacceptable
+956	acceptable
+957	acceptable
+958	acceptable
+959	acceptable
+960	acceptable
+961	acceptable
+962	acceptable
+963	acceptable
+964	acceptable
+965	acceptable
+966	acceptable
+967	acceptable
+968	acceptable
+969	acceptable
+970	acceptable
+971	acceptable
+972	acceptable
+973	acceptable
+974	acceptable
+975	acceptable
+976	acceptable
+977	acceptable
+978	acceptable
+979	acceptable
+980	acceptable
+981	unacceptable
+982	unacceptable
+983	acceptable
+984	acceptable
+985	unacceptable
+986	unacceptable
+987	acceptable
+988	unacceptable
+989	acceptable
+990	acceptable
+991	unacceptable
+992	acceptable
+993	unacceptable
+994	acceptable
+995	acceptable
+996	acceptable
+997	acceptable
+998	acceptable
+999	unacceptable
+1000	acceptable
+1001	acceptable
+1002	acceptable
+1003	acceptable
+1004	unacceptable
+1005	unacceptable
+1006	acceptable
+1007	unacceptable
+1008	acceptable
+1009	acceptable
+1010	acceptable
+1011	acceptable
+1012	unacceptable
+1013	acceptable
+1014	acceptable
+1015	acceptable
+1016	acceptable
+1017	acceptable
+1018	acceptable
+1019	acceptable
+1020	acceptable
+1021	acceptable
+1022	acceptable
+1023	unacceptable
+1024	acceptable
+1025	acceptable
+1026	acceptable
+1027	unacceptable
+1028	acceptable
+1029	acceptable
+1030	acceptable
+1031	acceptable
+1032	unacceptable
+1033	acceptable
+1034	acceptable
+1035	acceptable
+1036	unacceptable
+1037	acceptable
+1038	acceptable
+1039	acceptable
+1040	unacceptable
+1041	acceptable
+1042	acceptable
+1043	acceptable
+1044	unacceptable
+1045	unacceptable
+1046	unacceptable
+1047	acceptable
+1048	acceptable
+1049	unacceptable
+1050	unacceptable
+1051	unacceptable
+1052	unacceptable
+1053	unacceptable
+1054	unacceptable
+1055	unacceptable
+1056	acceptable
+1057	acceptable
+1058	acceptable
+1059	acceptable
+1060	acceptable
+1061	unacceptable
+1062	acceptable

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 16.0,
+    "train_loss": 0.18462801172133925,
+    "train_runtime": 361.7714,
+    "train_samples": 8551,
+    "train_samples_per_second": 378.184,
+    "train_steps_per_second": 2.963
+}

trainer_state.json

@@ -0,0 +1,1453 @@
+{
+  "best_metric": 0.5057060886900621,
+  "best_model_checkpoint": "./runtime-text-classification/paraphrase-MiniLM-L12-v2-CoLA/checkpoint-201",
+  "epoch": 16.0,
+  "global_step": 1072,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.2121212121212122e-05,
+      "loss": 0.691,
+      "step": 5
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2.4242424242424244e-05,
+      "loss": 0.6789,
+      "step": 10
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.6363636363636364e-05,
+      "loss": 0.6487,
+      "step": 15
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.848484848484849e-05,
+      "loss": 0.6241,
+      "step": 20
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.060606060606061e-05,
+      "loss": 0.6337,
+      "step": 25
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 7.272727272727273e-05,
+      "loss": 0.5772,
+      "step": 30
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.99992685960293e-05,
+      "loss": 0.6126,
+      "step": 35
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.99910406085351e-05,
+      "loss": 0.6145,
+      "step": 40
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.997367226544184e-05,
+      "loss": 0.5957,
+      "step": 45
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 7.994716753645944e-05,
+      "loss": 0.6062,
+      "step": 50
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 7.991153247951099e-05,
+      "loss": 0.6065,
+      "step": 55
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.986677523934811e-05,
+      "loss": 0.5945,
+      "step": 60
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.981290604568935e-05,
+      "loss": 0.5747,
+      "step": 65
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.539438784122467,
+      "eval_matthews_correlation": 0.34550903174440106,
+      "eval_runtime": 1.0831,
+      "eval_samples_per_second": 962.985,
+      "eval_steps_per_second": 60.937,
+      "step": 67
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 7.974993721088216e-05,
+      "loss": 0.5494,
+      "step": 70
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 7.96778831270887e-05,
+      "loss": 0.5121,
+      "step": 75
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 7.95967602629964e-05,
+      "loss": 0.5236,
+      "step": 80
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.95065871600539e-05,
+      "loss": 0.5188,
+      "step": 85
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 7.940738442823318e-05,
+      "loss": 0.5408,
+      "step": 90
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 7.929917474131895e-05,
+      "loss": 0.5426,
+      "step": 95
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 7.918198283172634e-05,
+      "loss": 0.5215,
+      "step": 100
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 7.905583548484803e-05,
+      "loss": 0.5201,
+      "step": 105
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 7.892076153293224e-05,
+      "loss": 0.5009,
+      "step": 110
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 7.877679184849275e-05,
+      "loss": 0.485,
+      "step": 115
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 7.86239593372527e-05,
+      "loss": 0.5062,
+      "step": 120
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.846229893062373e-05,
+      "loss": 0.5096,
+      "step": 125
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.829184757772192e-05,
+      "loss": 0.5025,
+      "step": 130
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.4999140501022339,
+      "eval_matthews_correlation": 0.42696977338948733,
+      "eval_runtime": 1.0932,
+      "eval_samples_per_second": 954.042,
+      "eval_steps_per_second": 60.371,
+      "step": 134
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.811264423692282e-05,
+      "loss": 0.4893,
+      "step": 135
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.792472986695704e-05,
+      "loss": 0.3651,
+      "step": 140
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.772814741754875e-05,
+      "loss": 0.4004,
+      "step": 145
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.752294181959911e-05,
+      "loss": 0.3846,
+      "step": 150
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.730915997491679e-05,
+      "loss": 0.3769,
+      "step": 155
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.708685074549819e-05,
+      "loss": 0.3736,
+      "step": 160
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.685606494235943e-05,
+      "loss": 0.3752,
+      "step": 165
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.661685531392307e-05,
+      "loss": 0.4007,
+      "step": 170
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.636927653396184e-05,
+      "loss": 0.3887,
+      "step": 175
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.611338518910249e-05,
+      "loss": 0.3895,
+      "step": 180
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.584923976589229e-05,
+      "loss": 0.3789,
+      "step": 185
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.557690063743136e-05,
+      "loss": 0.3876,
+      "step": 190
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.529643004957372e-05,
+      "loss": 0.3893,
+      "step": 195
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.500789210670052e-05,
+      "loss": 0.3698,
+      "step": 200
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.46361204981803894,
+      "eval_matthews_correlation": 0.5057060886900621,
+      "eval_runtime": 1.0755,
+      "eval_samples_per_second": 969.746,
+      "eval_steps_per_second": 61.365,
+      "step": 201
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 7.47113527570682e-05,
+      "loss": 0.3288,
+      "step": 205
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 7.44068797777354e-05,
+      "loss": 0.283,
+      "step": 210
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 7.409454275907184e-05,
+      "loss": 0.2778,
+      "step": 215
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 7.377441308885273e-05,
+      "loss": 0.2891,
+      "step": 220
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 7.344656393594239e-05,
+      "loss": 0.2637,
+      "step": 225
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 7.311107023357071e-05,
+      "loss": 0.2877,
+      "step": 230
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 7.27680086622065e-05,
+      "loss": 0.2744,
+      "step": 235
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 7.241745763203128e-05,
+      "loss": 0.2728,
+      "step": 240
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 7.205949726501796e-05,
+      "loss": 0.2808,
+      "step": 245
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 7.169420937661808e-05,
+      "loss": 0.3497,
+      "step": 250
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 7.132167745706208e-05,
+      "loss": 0.3107,
+      "step": 255
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 7.094198665227675e-05,
+      "loss": 0.2813,
+      "step": 260
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 7.055522374442433e-05,
+      "loss": 0.2969,
+      "step": 265
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 0.5308544039726257,
+      "eval_matthews_correlation": 0.4751366513569292,
+      "eval_runtime": 1.4776,
+      "eval_samples_per_second": 705.87,
+      "eval_steps_per_second": 44.667,
+      "step": 268
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 7.01614771320675e-05,
+      "loss": 0.3118,
+      "step": 270
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 6.976083680996507e-05,
+      "loss": 0.2078,
+      "step": 275
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 6.935339434850266e-05,
+      "loss": 0.1871,
+      "step": 280
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 6.89392428727635e-05,
+      "loss": 0.2217,
+      "step": 285
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 6.851847704124361e-05,
+      "loss": 0.2802,
+      "step": 290
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 6.809119302421682e-05,
+      "loss": 0.2718,
+      "step": 295
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 6.765748848175391e-05,
+      "loss": 0.2333,
+      "step": 300
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 6.721746254140152e-05,
+      "loss": 0.2585,
+      "step": 305
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 6.677121577552554e-05,
+      "loss": 0.2538,
+      "step": 310
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 6.631885017832421e-05,
+      "loss": 0.2754,
+      "step": 315
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 6.586046914251644e-05,
+      "loss": 0.2585,
+      "step": 320
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 6.539617743571023e-05,
+      "loss": 0.2141,
+      "step": 325
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 6.492608117645711e-05,
+      "loss": 0.2563,
+      "step": 330
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 6.445028780999751e-05,
+      "loss": 0.2275,
+      "step": 335
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.623831033706665,
+      "eval_matthews_correlation": 0.4774618517292431,
+      "eval_runtime": 1.0964,
+      "eval_samples_per_second": 951.284,
+      "eval_steps_per_second": 60.196,
+      "step": 335
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 6.396890608370326e-05,
+      "loss": 0.1801,
+      "step": 340
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 6.348204602222213e-05,
+      "loss": 0.1687,
+      "step": 345
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 6.298981890233062e-05,
+      "loss": 0.1548,
+      "step": 350
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 6.24923372275006e-05,
+      "loss": 0.1502,
+      "step": 355
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 6.198971470218543e-05,
+      "loss": 0.1816,
+      "step": 360
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 6.148206620583172e-05,
+      "loss": 0.1695,
+      "step": 365
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 6.096950776662239e-05,
+      "loss": 0.1828,
+      "step": 370
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 6.0452156534957336e-05,
+      "loss": 0.1768,
+      "step": 375
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 5.993013075667753e-05,
+      "loss": 0.2368,
+      "step": 380
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 5.940354974603868e-05,
+      "loss": 0.1782,
+      "step": 385
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 5.887253385844076e-05,
+      "loss": 0.1508,
+      "step": 390
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 5.8337204462919645e-05,
+      "loss": 0.1778,
+      "step": 395
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 5.779768391440693e-05,
+      "loss": 0.1859,
+      "step": 400
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 0.6314846873283386,
+      "eval_matthews_correlation": 0.4867092923459376,
+      "eval_runtime": 1.0756,
+      "eval_samples_per_second": 969.736,
+      "eval_steps_per_second": 61.364,
+      "step": 402
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 5.725409552576445e-05,
+      "loss": 0.1603,
+      "step": 405
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 5.670656353959997e-05,
+      "loss": 0.1204,
+      "step": 410
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 5.615521309987022e-05,
+      "loss": 0.1699,
+      "step": 415
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 5.5600170223278e-05,
+      "loss": 0.1502,
+      "step": 420
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 5.5041561770469787e-05,
+      "loss": 0.1389,
+      "step": 425
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 5.4479515417040426e-05,
+      "loss": 0.1567,
+      "step": 430
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 5.391415962435167e-05,
+      "loss": 0.1644,
+      "step": 435
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 5.334562361017097e-05,
+      "loss": 0.1748,
+      "step": 440
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 5.277403731913743e-05,
+      "loss": 0.1308,
+      "step": 445
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 5.2199531393061755e-05,
+      "loss": 0.1034,
+      "step": 450
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 5.162223714106661e-05,
+      "loss": 0.1266,
+      "step": 455
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 5.104228650957472e-05,
+      "loss": 0.0935,
+      "step": 460
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 5.0459812052151114e-05,
+      "loss": 0.1517,
+      "step": 465
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 0.7783187627792358,
+      "eval_matthews_correlation": 0.4695101101903524,
+      "eval_runtime": 1.0697,
+      "eval_samples_per_second": 975.006,
+      "eval_steps_per_second": 61.697,
+      "step": 469
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.9874946899206665e-05,
+      "loss": 0.1364,
+      "step": 470
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.928782472756985e-05,
+      "loss": 0.1386,
+      "step": 475
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 4.869857972993352e-05,
+      "loss": 0.0828,
+      "step": 480
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.810734658418377e-05,
+      "loss": 0.123,
+      "step": 485
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.751426042261806e-05,
+      "loss": 0.0893,
+      "step": 490
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.691945680105929e-05,
+      "loss": 0.108,
+      "step": 495
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.63230716678731e-05,
+      "loss": 0.0797,
+      "step": 500
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.5725241332895643e-05,
+      "loss": 0.1078,
+      "step": 505
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.512610243627845e-05,
+      "loss": 0.1057,
+      "step": 510
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.4525791917258025e-05,
+      "loss": 0.0968,
+      "step": 515
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.3924446982857027e-05,
+      "loss": 0.1196,
+      "step": 520
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.3322205076524145e-05,
+      "loss": 0.1423,
+      "step": 525
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.271920384672002e-05,
+      "loss": 0.1203,
+      "step": 530
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.21155811154564e-05,
+      "loss": 0.1016,
+      "step": 535
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 0.6762046813964844,
+      "eval_matthews_correlation": 0.49007739162498226,
+      "eval_runtime": 1.0748,
+      "eval_samples_per_second": 970.453,
+      "eval_steps_per_second": 61.409,
+      "step": 536
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.151147484679541e-05,
+      "loss": 0.0866,
+      "step": 540
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.0907023115316525e-05,
+      "loss": 0.0947,
+      "step": 545
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 4.030236407455827e-05,
+      "loss": 0.0718,
+      "step": 550
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 3.969763592544173e-05,
+      "loss": 0.0729,
+      "step": 555
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 3.909297688468347e-05,
+      "loss": 0.1148,
+      "step": 560
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 3.84885251532046e-05,
+      "loss": 0.1083,
+      "step": 565
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 3.7884418884543614e-05,
+      "loss": 0.1135,
+      "step": 570
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 3.728079615327999e-05,
+      "loss": 0.0581,
+      "step": 575
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 3.667779492347588e-05,
+      "loss": 0.0943,
+      "step": 580
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.607555301714297e-05,
+      "loss": 0.095,
+      "step": 585
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.5474208082741975e-05,
+      "loss": 0.1263,
+      "step": 590
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.4873897563721565e-05,
+      "loss": 0.1371,
+      "step": 595
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.427475866710437e-05,
+      "loss": 0.1017,
+      "step": 600
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 0.741180419921875,
+      "eval_matthews_correlation": 0.5046330570516372,
+      "eval_runtime": 1.0744,
+      "eval_samples_per_second": 970.819,
+      "eval_steps_per_second": 61.432,
+      "step": 603
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 3.367692833212691e-05,
+      "loss": 0.0929,
+      "step": 605
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 3.308054319894074e-05,
+      "loss": 0.0638,
+      "step": 610
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 3.2485739577381946e-05,
+      "loss": 0.1338,
+      "step": 615
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 3.1892653415816245e-05,
+      "loss": 0.0663,
+      "step": 620
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 3.13014202700665e-05,
+      "loss": 0.0767,
+      "step": 625
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 3.071217527243016e-05,
+      "loss": 0.1068,
+      "step": 630
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 3.012505310079335e-05,
+      "loss": 0.0818,
+      "step": 635
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.9540187947848906e-05,
+      "loss": 0.0527,
+      "step": 640
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.8957713490425298e-05,
+      "loss": 0.0837,
+      "step": 645
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.837776285893339e-05,
+      "loss": 0.0804,
+      "step": 650
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.780046860693826e-05,
+      "loss": 0.1138,
+      "step": 655
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.7225962680862575e-05,
+      "loss": 0.0705,
+      "step": 660
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.6654376389829043e-05,
+      "loss": 0.0562,
+      "step": 665
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.6085840375648337e-05,
+      "loss": 0.0898,
+      "step": 670
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.7719386219978333,
+      "eval_matthews_correlation": 0.48769484136865787,
+      "eval_runtime": 1.1003,
+      "eval_samples_per_second": 947.882,
+      "eval_steps_per_second": 59.981,
+      "step": 670
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.5520484582959588e-05,
+      "loss": 0.0739,
+      "step": 675
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4958438229530223e-05,
+      "loss": 0.0927,
+      "step": 680
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4399829776722005e-05,
+      "loss": 0.0731,
+      "step": 685
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.384478690012978e-05,
+      "loss": 0.051,
+      "step": 690
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.3293436460400044e-05,
+      "loss": 0.0647,
+      "step": 695
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.2745904474235573e-05,
+      "loss": 0.0704,
+      "step": 700
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.2202316085593106e-05,
+      "loss": 0.0479,
+      "step": 705
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.1662795537080365e-05,
+      "loss": 0.076,
+      "step": 710
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.1127466141559257e-05,
+      "loss": 0.0709,
+      "step": 715
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.059645025396135e-05,
+      "loss": 0.0698,
+      "step": 720
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.006986924332247e-05,
+      "loss": 0.0606,
+      "step": 725
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 1.9547843465042664e-05,
+      "loss": 0.096,
+      "step": 730
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 1.9030492233377623e-05,
+      "loss": 0.0527,
+      "step": 735
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 0.8626514673233032,
+      "eval_matthews_correlation": 0.49547617383910947,
+      "eval_runtime": 1.0828,
+      "eval_samples_per_second": 963.257,
+      "eval_steps_per_second": 60.954,
+      "step": 737
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 1.8517933794168294e-05,
+      "loss": 0.0582,
+      "step": 740
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 1.8010285297814563e-05,
+      "loss": 0.0731,
+      "step": 745
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 1.7507662772499406e-05,
+      "loss": 0.0478,
+      "step": 750
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 1.701018109766939e-05,
+      "loss": 0.0564,
+      "step": 755
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 1.6517953977777888e-05,
+      "loss": 0.0413,
+      "step": 760
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 1.603109391629675e-05,
+      "loss": 0.0478,
+      "step": 765
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 1.5549712190002498e-05,
+      "loss": 0.0844,
+      "step": 770
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 1.50739188235429e-05,
+      "loss": 0.065,
+      "step": 775
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 1.4603822564289764e-05,
+      "loss": 0.068,
+      "step": 780
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 1.4139530857483568e-05,
+      "loss": 0.0324,
+      "step": 785
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 1.3681149821675797e-05,
+      "loss": 0.0436,
+      "step": 790
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 1.322878422447448e-05,
+      "loss": 0.0566,
+      "step": 795
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 1.2782537458598489e-05,
+      "loss": 0.0582,
+      "step": 800
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 0.8985734581947327,
+      "eval_matthews_correlation": 0.47375271850377065,
+      "eval_runtime": 1.0831,
+      "eval_samples_per_second": 962.944,
+      "eval_steps_per_second": 60.934,
+      "step": 804
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 1.2342511518246095e-05,
+      "loss": 0.0521,
+      "step": 805
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.1908806975783192e-05,
+      "loss": 0.0565,
+      "step": 810
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.1481522958756383e-05,
+      "loss": 0.0358,
+      "step": 815
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.1060757127236514e-05,
+      "loss": 0.0436,
+      "step": 820
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.0646605651497342e-05,
+      "loss": 0.06,
+      "step": 825
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.0239163190034937e-05,
+      "loss": 0.0439,
+      "step": 830
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 9.838522867932499e-06,
+      "loss": 0.0436,
+      "step": 835
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 9.444776255575672e-06,
+      "loss": 0.032,
+      "step": 840
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 9.058013347723258e-06,
+      "loss": 0.0623,
+      "step": 845
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 8.678322542937941e-06,
+      "loss": 0.0526,
+      "step": 850
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 8.305790623381934e-06,
+      "loss": 0.0231,
+      "step": 855
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 7.94050273498205e-06,
+      "loss": 0.035,
+      "step": 860
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 7.582542367968733e-06,
+      "loss": 0.0612,
+      "step": 865
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 7.2319913377935096e-06,
+      "loss": 0.074,
+      "step": 870
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 0.9468681812286377,
+      "eval_matthews_correlation": 0.49419533040588925,
+      "eval_runtime": 1.0698,
+      "eval_samples_per_second": 974.98,
+      "eval_steps_per_second": 61.696,
+      "step": 871
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 6.888929766429293e-06,
+      "loss": 0.0266,
+      "step": 875
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 6.553436064057619e-06,
+      "loss": 0.0631,
+      "step": 880
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 6.225586911147274e-06,
+      "loss": 0.0549,
+      "step": 885
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 5.9054572409281655e-06,
+      "loss": 0.0212,
+      "step": 890
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 5.5931202222646145e-06,
+      "loss": 0.0404,
+      "step": 895
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 5.288647242931815e-06,
+      "loss": 0.0606,
+      "step": 900
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 4.992107893299483e-06,
+      "loss": 0.0523,
+      "step": 905
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 4.703569950426285e-06,
+      "loss": 0.0283,
+      "step": 910
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 4.4230993625686616e-06,
+      "loss": 0.0515,
+      "step": 915
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 4.150760234107716e-06,
+      "loss": 0.0333,
+      "step": 920
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 3.88661481089752e-06,
+      "loss": 0.0338,
+      "step": 925
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 3.630723466038175e-06,
+      "loss": 0.0344,
+      "step": 930
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 3.3831446860769444e-06,
+      "loss": 0.0508,
+      "step": 935
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": 0.9435966610908508,
+      "eval_matthews_correlation": 0.49177365217771035,
+      "eval_runtime": 1.082,
+      "eval_samples_per_second": 963.975,
+      "eval_steps_per_second": 60.999,
+      "step": 938
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 3.1439350576405726e-06,
+      "loss": 0.0513,
+      "step": 940
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.913149254501817e-06,
+      "loss": 0.0477,
+      "step": 945
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.69084002508321e-06,
+      "loss": 0.0403,
+      "step": 950
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.477058180400902e-06,
+      "loss": 0.0657,
+      "step": 955
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.2718525824512614e-06,
+      "loss": 0.0454,
+      "step": 960
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.0752701330429702e-06,
+      "loss": 0.0314,
+      "step": 965
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.887355763077192e-06,
+      "loss": 0.0261,
+      "step": 970
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.7081524222780864e-06,
+      "loss": 0.0366,
+      "step": 975
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.5377010693762784e-06,
+      "loss": 0.0307,
+      "step": 980
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3760406627473022e-06,
+      "loss": 0.0604,
+      "step": 985
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.2232081515072624e-06,
+      "loss": 0.0477,
+      "step": 990
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.0792384670677625e-06,
+      "loss": 0.0303,
+      "step": 995
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 9.441645151519708e-07,
+      "loss": 0.032,
+      "step": 1000
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 8.180171682736682e-07,
+      "loss": 0.024,
+      "step": 1005
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": 0.9390689730644226,
+      "eval_matthews_correlation": 0.49194159281920924,
+      "eval_runtime": 1.0846,
+      "eval_samples_per_second": 961.661,
+      "eval_steps_per_second": 60.853,
+      "step": 1005
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 7.008252586810571e-07,
+      "loss": 0.0518,
+      "step": 1010
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 5.92615571766828e-07,
+      "loss": 0.0266,
+      "step": 1015
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 4.934128399461057e-07,
+      "loss": 0.0424,
+      "step": 1020
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 4.032397370036067e-07,
+      "loss": 0.0441,
+      "step": 1025
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 3.221168729113089e-07,
+      "loss": 0.0388,
+      "step": 1030
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 2.5006278911784643e-07,
+      "loss": 0.0486,
+      "step": 1035
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.8709395431065535e-07,
+      "loss": 0.0311,
+      "step": 1040
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.3322476065190438e-07,
+      "loss": 0.023,
+      "step": 1045
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 8.846752048901952e-08,
+      "loss": 0.0396,
+      "step": 1050
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 5.2832463540575254e-08,
+      "loss": 0.0288,
+      "step": 1055
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 2.6327734558173613e-08,
+      "loss": 0.028,
+      "step": 1060
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 8.95939146489777e-09,
+      "loss": 0.0409,
+      "step": 1065
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 7.314039707040366e-10,
+      "loss": 0.0458,
+      "step": 1070
+    },
+    {
+      "epoch": 16.0,
+      "eval_loss": 0.9374607801437378,
+      "eval_matthews_correlation": 0.49464326454019025,
+      "eval_runtime": 1.0777,
+      "eval_samples_per_second": 967.817,
+      "eval_steps_per_second": 61.243,
+      "step": 1072
+    },
+    {
+      "epoch": 16.0,
+      "step": 1072,
+      "total_flos": 2253110874603520.0,
+      "train_loss": 0.18462801172133925,
+      "train_runtime": 361.7714,
+      "train_samples_per_second": 378.184,
+      "train_steps_per_second": 2.963
+    }
+  ],
+  "max_steps": 1072,
+  "num_train_epochs": 16,
+  "total_flos": 2253110874603520.0,
+  "trial_name": null,
+  "trial_params": null
+}