MCPcopy Index your code
hub / github.com/zai-org/CogView / load_weights

Function load_weights

utils.py:383–400  ·  view source on GitHub ↗

Loads weights from src to dst via in place copy. src is a huggingface gpt2model, while dst is one of our models. dst2src=True loads parameters from our models into huggingface's. ^dst2src is still untested

(src, dst, dst2src=False)

Source from the content-addressed store, hash-verified

381
382
383def load_weights(src, dst, dst2src=False):
384 """
385 Loads weights from src to dst via in place copy.
386 src is a huggingface gpt2model, while dst is one of our models.
387 dst2src=True loads parameters from our models into huggingface's.
388 ^dst2src is still untested
389 """
390 conv_layer = 'Conv1D' in str(type(src))
391 for n, p in src.named_parameters():
392 if dst2src:
393 data = dst._parameters[n].data
394 load = p.data
395 else:
396 data = p.data
397 load = dst._parameters[n].data
398 if conv_layer and 'weight' in n:
399 data = data.t().contiguous()
400 load.copy_(data)
401
402
403# dst._parameters[n].data.copy_(data)

Callers 4

load_mlpFunction · 0.85
load_attentionFunction · 0.85
load_transformer_layerFunction · 0.85
move_weightsFunction · 0.85

Calls

no outgoing calls

Tested by

no test coverage detected