MCPcopy Create free account
hub / github.com/deepspeedai/DeepSpeedExamples / load_mlp

Function load_mlp

Megatron-LM/utils.py:381–383  ·  view source on GitHub ↗
(our, oai, dst2src=False)

Source from the content-addressed store, hash-verified

379# dst._parameters[n].data.copy_(data)
380
381def load_mlp(our, oai, dst2src=False):
382 load_weights(oai.c_fc, our.dense_h_to_4h, dst2src)
383 load_weights(oai.c_proj, our.dense_4h_to_h, dst2src)
384
385def load_attention(our, oai, dst2src=False):
386 load_weights(oai.c_attn, our.query_key_value, dst2src)

Callers 1

load_transformer_layerFunction · 0.85

Calls 1

load_weightsFunction · 0.85

Tested by

no test coverage detected