Which models are very accurate in general for forecasting covid 19 infection cases (ARIMA, Holt's Linear Trend, BATS, TBATS, SIR, NNAR, LSTM, BiLSTM, GRU).
There is no ideal model for forecasting COVID-19 infection cases. Also, all models aren't suitable in all cases. So, choosing the best model is a tough task. However, you should try to look for all possible [time varying reproduction number (based on Bayesian approach), SIR, SRS, SEIR, SEIRM, etc.] models and related errors (if any). Errors will guide you to select a better model, but not the best. Hope, this may help.