make psychometrics curve fitting more robust

lms/djangoapps/psychometrics/psychoanalyze.py

@@ -246,6 +246,7 @@ def generate_plots_for_problem(problem):
         yset['ydat'] = ydat
 
         if len(ydat) > 3:         # try to fit to logistic function if enough data points
+            try:
                 cfp = curve_fit(func_2pl, xdat, ydat, [1.0, max_attempts / 2.0])
                 yset['fitparam'] = cfp
                 yset['fitpts'] = func_2pl(np.array(xdat), *cfp[0])
@@ -253,6 +254,8 @@ def generate_plots_for_problem(problem):
                 fitx = np.linspace(xdat[0], xdat[-1], 100)
                 yset['fitx'] = fitx
                 yset['fity'] = func_2pl(np.array(fitx), *cfp[0])
+            except Exception as err:
+                log.debug('Error in psychoanalyze curve fitting: %s' % err)
 
         dataset['grade_%d' % grade] = yset