leave separate memory attention parameters for another day

pi_zero_pytorch/pi_zero.py

@@ -53,9 +53,11 @@
 # vision and language tokens are autoregressive causal mask, actions, interal states + joint bidirectional amongst own tokens, but still autoregressive with respect to other tokens
 
 # [state token groups] [action token groups] -> [autoregressive masking] [bidirectional]
-
 # [external state] [visual tokens] [language tokens] [maybe reward / condition token] [action registers] [joint state + internal state] [actions]
 
+# for an attempt to introduce recurrence, all tokens above can be flanked by read and write memory tokens
+# [read memory tokens] [...] [write memory tokens]
+
 # constants
 
 LinearNoBias = partial(nn.Linear, bias = False)