CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-03 09:22:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : add Deepseek MoE v1 & GigaChat models (#10827)

Browse Source 
* Add deepseek v1 arch & gigachat template

* improve template code

* add readme

* delete comments

* remove comment

* fix format

* lint llama.cpp

* fix order of deepseek and deepseek2, move gigachat temlate to the end of func

* fix order of deepseek and deepseek2 in constants; mark shared exp as deepseek arch need

* remove comments

* move deepseek above deepseek2

* change placement of gigachat chat template
This commit is contained in:
Valentin Mamedov
2024-12-16 00:02:46 +07:00
committed by GitHub
parent 87cf323cef
commit a0974156f3
7 changed files with 423 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
gguf-py/gguf/tensor_mapping.py
View File

@@ -306,7 +306,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_UP_SHEXP: (
"model.layers.{bid}.mlp.shared_expert.up_proj", # qwen2moe
"model.layers.{bid}.mlp.shared_experts.up_proj", # deepseek2
"model.layers.{bid}.mlp.shared_experts.up_proj", # deepseek deepseek2
),
# AWQ-activation gate
@@ -338,7 +338,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_GATE_SHEXP: (
"model.layers.{bid}.mlp.shared_expert.gate_proj", # qwen2moe
"model.layers.{bid}.mlp.shared_experts.gate_proj", # deepseek2
"model.layers.{bid}.mlp.shared_experts.gate_proj", # deepseek deepseek2
),
# Feed-forward down
@@ -379,7 +379,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_DOWN_SHEXP: (
"model.layers.{bid}.mlp.shared_expert.down_proj", # qwen2moe
"model.layers.{bid}.mlp.shared_experts.down_proj", # deepseek2
"model.layers.{bid}.mlp.shared_experts.down_proj", # deepseek deepseek2
),
MODEL_TENSOR.ATTN_Q_NORM: (