MCPcopy Create free account
hub / github.com/EmbodiedGPT/EmbodiedGPT_Pytorch / forward

Method forward

robohusky/compression.py:45–49  ·  view source on GitHub ↗
(self, input: Tensor)

Source from the content-addressed store, hash-verified

43 self.bias = bias
44
45 def forward(self, input: Tensor) -> Tensor:
46 weight = decompress(self.weight, default_compression_config)
47 if self.bias is None:
48 return F.linear(input.to(weight.dtype), weight)
49 return F.linear(input.to(weight.dtype), weight, self.bias.to(weight.dtype))
50
51
52def compress_module(module, target_device):

Callers 1

testFunction · 0.45

Calls 1

decompressFunction · 0.85

Tested by

no test coverage detected