vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

convert: dbrx: fix mixed up and down expert tensors

Browse source
This commit is contained in:
Pierrick HYMBERT 2024-04-06 19:21:43 +02:00
parent a7f9a3eafc
commit e3c1e8127c
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
gguf-py/gguf/tensor_mapping.py
View file

@ -238,7 +238,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_UP_EXP: (
"layers.{bid}.feed_forward.experts.w3", # mixtral (merged)
"transformer.decoder_layer.{bid}.moe.linear_v", # Grok (merged)
"transformer.blocks.{bid}.ffn.experts.mlp.w2", # dbrx
"transformer.blocks.{bid}.ffn.experts.mlp.w1", # dbrx
),
# AWQ-activation gate
@ -287,7 +287,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_DOWN_EXP: (
"layers.{bid}.feed_forward.experts.w2", # mixtral (merged)
"transformer.decoder_layer.{bid}.moe.linear_1", # Grok (merged)
"transformer.blocks.{bid}.ffn.experts.mlp.w1", # dbrx
"transformer.blocks.{bid}.ffn.experts.mlp.w2", # dbrx
),
MODEL_TENSOR.ATTN_Q_NORM: (