Merge pull request #85 from VE-FORBRYDERNE/sp
Fix a bug with soft prompts when using transformers XGLM
This commit is contained in:
commit
f628929401
|
@ -1048,7 +1048,7 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly", "TPUMeshTransforme
|
|||
if(hasattr(self, "transformer")):
|
||||
inputs_embeds = self.transformer.wte(input_ids)
|
||||
else:
|
||||
inputs_embeds = self.model.embed_tokens(input_ids) * self.model.embed_scale
|
||||
inputs_embeds = self.model.embed_tokens(input_ids)
|
||||
if(vars.sp is not None):
|
||||
vars.sp = vars.sp.to(inputs_embeds.dtype).to(inputs_embeds.device)
|
||||
inputs_embeds = torch.where(
|
||||
|
@ -1056,6 +1056,8 @@ if(not vars.model in ["InferKit", "Colab", "OAI", "ReadOnly", "TPUMeshTransforme
|
|||
vars.sp[shifted_input_ids.clamp(min=0)],
|
||||
inputs_embeds,
|
||||
)
|
||||
if(not hasattr(self, "transformer")):
|
||||
inputs_embeds *= self.model.embed_scale
|
||||
kwargs['inputs_embeds'] = inputs_embeds
|
||||
return old_forward(self, *args, **kwargs)
|
||||
cls.forward = new_causallm_forward
|
||||
|
|
Loading…
Reference in New Issue