Bokeh serve app plot incomplet

hi,
I’m new to bokeh .
now I found that the bokeh serve app plot incomplet when useing the serve mode
the code plot the three figure dynamic using webgl mode. I found that the figure is incomplet 。the title ,the label of yaxis,label of xaxis not shown 。 but it is complet sometimes so it is puzzling
system bokeh 1.4, python 3.7,chrome browser ,win 10
Thanks so much!
the code show below

pelb=BoxAnnotation(top=dpe_ttm['20%'],fill_alpha=0.1,fill_color='green')
pemb=BoxAnnotation(bottom=dpe_ttm['40%'],top=dpe_ttm['60%'],fill_alpha=0.1,fill_color='coral')
pemg=BoxAnnotation(bottom=dpe_ttm['75%'],top=dpe_ttm['max'],fill_alpha=0.1,fill_color='crimson')
gp_p1.add_layout(pelb)
gp_p1.add_layout(pemb)
gp_p1.add_layout(pemg)

gp_p1.xaxis.axis_label='日期'
gp_p1.yaxis.axis_label='市盈率'
gp_p1.title.text=code+'市盈率数据'

gp_p1.line(x='trade_date',y=dpe['max'],source=tc,color='orangered',legend_label='PE 最大值')
gp_p1.line(x='trade_date',y=dpe['min'],source=tc,color='navy',legend_label='PE 最小值')

gp_p1.line(x='trade_date',y='pe',source=tc,\
           color='blue',legend_label='市盈率')
gp_p1.line(x='trade_date',y='pe_ttm',source=tc,\
            color='red',legend_label='滚动市盈率')

gp_p1.add_tools(HoverTool(tooltips=[("日期", '@trade_date{%F}'), ("市盈率", "@pe"),\
                                    ("滚动市盈率", "@pe_ttm")],\
                formatters={'trade_date':'datetime'},mode='vline'))


gp_p1.legend.click_policy='hide'

deng

Hi @wwwcisco There is nothing immediately amiss just from inspecting the code snippet above. Really, in order to investigate, we would need a minimal, complete reproducer that can be run to investigate. An MRE is a self-contained toy example that has just enough code to reproduce the problem you are seeing without anything extraneous or unrelated to the issue at hand.

hi,thanks for reply ,can you give me a mail address ,so I can share the all code or the website which is private

@wwwcisco you should be able to send me a private message here on there forum, by clicking on my name in the message here, Another option is always to make a minimal, toy example, with fake data, that illustrates the same problem.