Following is my code to the problem Link

I=lambda:[*map(int,input().split())]
rr=range
n,m=I()
a=[I() for _ in rr(n)]
# print(a)
dp=[0]*(n+1)
for i in rr(m):
    st=en=0
    for j in rr(1,n):
        if a[j-1][i]>a[j][i]:
            dp[st+1]=max(dp[st+1],en+1)
            st=en=j
        else :
            en+=1
    dp[st+1]=max(dp[st+1],en+1)
for i in rr(1,n+1):
    dp[i]=max(dp[i],dp[i-1])

q,=I()
for i in rr(q):
    x,y=I()
    print('Yes' if dp[x]>=y else 'No')

It gives TLE in some submissions while passes in others, to the fact that all have same code. Also sometimes for the same code Python 3 throws TLE and sometimes PyPy3. Is there any way to overcome this uncertainty, Or can anyone pls give reason for this