fix bugs in documentation

Author: qiauil

docs/start/theory.md

@@ -73,15 +73,15 @@ $$
 
 where $\mathcal{O}(\mathbf{g}_1,\mathbf{g}_2)=\mathbf{g}_2-\frac{\mathbf{g}_1^\top\mathbf{g}_2}{|\mathbf{g}1|^2}\mathbf{g}_1$ is the orthogonality operator. It returns a vector orthogonal to $\mathbf{g}_1$ from the plane spanned by $\mathbf{g}_{1}$ and $\mathbf{g}_{2}$. 
 
-This equivlance is implemented as [ConFIG.grad_operator.ConFIG_update_double()](../../api/grad_operator/#ConFIG.grad_operator.ConFIG_update_double. You can also set `allow_simplified_model` to true in [ConFIG.grad_operator.ConFIGOperator](../../api/grad_operator/#ConFIG.grad_operator.ConFIGOperator) to enable using this form in two-loss scenario.
+This equivlance is implemented as [ConFIG.grad_operator.ConFIG_update_double()](../../api/grad_operator/#ConFIG.grad_operator.ConFIG_update_double). You can also set `allow_simplified_model` to true in [ConFIG.grad_operator.ConFIGOperator](../../api/grad_operator/#ConFIG.grad_operator.ConFIGOperator) to enable using this form in two-loss scenario.
 
 ## M-ConFIG
 
 Gradient-based methods like the proposed ConFIG method require separate backpropagation steps to compute the gradient for each loss term, which could be computationally expensive.  To address this issue, we introduce an accelerated momentum-based variant of ConFIG: **M-ConFIG**. Our core idea is to leverage the momentum of the gradient for the ConFIG operation and update momentum variables in an alternating fashion to avoid backpropagating all losses in a single step. In each iteration, only a single momentum is updated with its corresponding gradient, while the others are carried over from previous steps.  Algorithm 1 details the entire procedure of M-ConFIG.
 
 ![M-ConFIG](../assets/algorithm.png)
 
-The M-ConFIG method is implemented as [ConFIG.momentum_operator.PseudoMomentumOperator](../../api/momentum_operator/#ConFIG.momentum_operator.PseudoMomentumOperator). This momentum method can also be used for other gradient-based methods. In [ConFIG.momentum_operator.PseudoMomentumOperator](../../api/momentum_operator/#ConFIG.momentum_operator.PseudoMomentumOperator),you can modify the `gradient_operator` parameter to enable momentum acceleration for other methods.
+The M-ConFIG method is implemented as [ConFIG.momentum_operator.PseudoMomentumOperator](../../api/momentum_operator/#ConFIG.momentum_operator.PseudoMomentumOperator). This momentum method can also be used for other gradient-based methods. In [ConFIG.momentum_operator.PseudoMomentumOperator](../../api/momentum_operator/#ConFIG.momentum_operator.PseudoMomentumOperator), you can modify the `gradient_operator` parameter to enable momentum acceleration for other methods.
 
 ---