[LLM] Reconstruct fused transformer layers

[RichardWooSJTU]

PR types

Function optimization

PR changes

Others

Description

Reconstruct class FusedMultiTransformer

• Add FusedMultiTransformerConfig
• Add FusedMultiTransformerBase
• Add FusedMultiTransformerWeightonly
• Add FusedMultiTransformerPostLayernorm
• Add FusedMultiTransformerWeightonlyPostLayernorm

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Merging #7186 (c3dcf12) into develop (74eb855) will decrease coverage by 0.06%.
The diff coverage is 0.00%.

@@             Coverage Diff             @@
##           develop    #7186      +/-   ##
===========================================
- Coverage    59.39%   59.33%   -0.06%     
===========================================
  Files          567      567              
  Lines        83114    83190      +76     
===========================================
  Hits         49364    49364              
- Misses       33750    33826      +76     

Files	Coverage Δ
...dlenlp/experimental/transformers/bloom/modeling.py	0.00% <0.00%> (ø)
...enlp/experimental/transformers/chatglm/modeling.py	0.00% <0.00%> (ø)
...addlenlp/experimental/transformers/gpt/modeling.py	0.00% <0.00%> (ø)
...addlenlp/experimental/transformers/opt/modeling.py	0.00% <0.00%> (ø)
...dlenlp/experimental/transformers/llama/modeling.py	0.00% <0.00%> (ø)
...erimental/transformers/fused_transformer_layers.py	0.00% <0.00%> (ø)

[wj-Mcat]

这个PR 导致之前的推理单测挂了：https://xly.bce.baidu.com/paddlepaddle/Paddle-NLP/newipipe/detail/9304392/job/24091206

这样看来可能会影响之前模型的精度。

[wj-Mcat]

合入 #7187 之后你这边有冲突，麻烦解决一下冲突。

[wj-Mcat]

LGTM