chatlearn/models/megatron/lora/layers.py [259:287]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.lora_scaling = lora_scaling / lora_dim

        if lora_dropout > 0:
            self.lora_dropout = nn.Dropout(lora_dropout)
        else:
            self.lora_dropout = nn.Identity()

        self.reset_parameters()
        # disable the original weight gradient
        self.weight.requires_grad = False
        # fuse LoRA to the original weight
        self.fuse_lora = False

    def eval(self):
        self.lora_dropout.eval()

    #   self.fuse_lora_weight()

    def train(self, mode=True):
        self.lora_dropout.train(mode)
        # self.unfuse_lora_weight()

    def reset_parameters(self):
        distributed_kaiming_uniform_(self.lora_right_weight, self.fan_in, a=math.sqrt(5))
        nn.init.zeros_(self.lora_left_weight)

    def fuse_lora_weight(self):
        if not self.fuse_lora:
            self.weight.data += self.lora_scaling * torch.matmul(
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



chatlearn/models/megatron/lora/layers.py [479:507]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        self.lora_scaling = lora_scaling / lora_dim

        if lora_dropout > 0:
            self.lora_dropout = nn.Dropout(lora_dropout)
        else:
            self.lora_dropout = nn.Identity()

        self.reset_parameters()
        # disable the original weight gradient
        self.weight.requires_grad = False
        # fuse LoRA to the original weight
        self.fuse_lora = False

    def eval(self):
        self.lora_dropout.eval()

    #   self.fuse_lora_weight()

    def train(self, mode=True):
        self.lora_dropout.train(mode)
        # self.unfuse_lora_weight()

    def reset_parameters(self):
        distributed_kaiming_uniform_(self.lora_right_weight, self.fan_in, a=math.sqrt(5))
        nn.init.zeros_(self.lora_left_weight)

    def fuse_lora_weight(self):
        if not self.fuse_lora:
            self.weight.data += self.lora_scaling * torch.matmul(
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -