Method forward

distributed/FSDP2/model.py:86–89

(self, x)

        )

    def forward(self, x):
        h = x + self.attention(self.attention_norm(x))
        out = h + self.feed_forward(self.ffn_norm(h))
        return out

    def reset_parameters(self):
        self.attention_norm.reset_parameters()