Update README.md

Signed-off-by: Rui Campos <mail@ruicampos.org>

README.md

@@ -62,7 +62,7 @@ $$q^{nul} _ {l} = \delta_{f(l)g(l)} \bar M^n_{l} \left [ \delta^{f(u)g(u)} p^{nf
 
 which we'll now group according to the $\delta$'s
 
-$$q^{nul} _ {l} =  \bar M^n _ {l} p^{nf(u)f(l)} p^{nf(u)f(l)} \delta_{f(l)g(l)} \delta^{f(u)g(u)}  +   2 \bar M^n_{l}  p^{nf(u)f(l)} p^{ng(u)f(l)} \delta_{f(l)g(l)} \tilde \delta^{f(u)g(u)} +    2    \bar M^n_l p^{nf(u)f(l)} p^{nf(u)g(l)} \delta^{f(u)g(u)} \tilde \delta_{f(l)g(l)} + 4   \bar M^n_l   p^{nf(u)f(l)} p^{ng(u)g(l)} \tilde \delta^{f(u)g(u)} \tilde \delta_{f(l)g(l)}$$
+$$q^{nul} _ {l} =  \bar M^n _ {l} p^{nf(u)f(l)} p^{nf(u)f(l)} \delta^{f(l)g(l)} \delta^{f(u)g(u)}  +   2 \bar M^n_{l}  p^{nf(u)f(l)} p^{ng(u)f(l)} \delta^{f(l)g(l)} \tilde \delta^{f(u)g(u)} +    2    \bar M^n_l p^{nf(u)f(l)} p^{nf(u)g(l)} \delta^{f(u)g(u)} \tilde \delta^{f(l)g(l)} + 4   \bar M^n_l   p^{nf(u)f(l)} p^{ng(u)g(l)} \tilde \delta^{f(u)g(u)} \tilde \delta^{f(l)g(l)}$$
 
 For a given combination of $u$ and $l$, there's only one term to be calculated. All terms will be computed in paralel on the gpu and collected onto a tensor that represents $q^{nul} _ {l}$, as demanded by the tensor notation, a sum is then performed over $l$ to obtain $q^{nu}$. The lookup tables for $f$ and $g$ are then used to recover $q^{ncc'}$ which is then comunicated back to torch for the rest of the attention mechanism.