EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

llama : support LiquidAI LFM2-MoE hybrid model (#16464)

Browse Source 
* llama : support LiquidAI LFM2-MoE hybrid model

Add support for [LiquidAI/LFM2-8B-A1B](https://huggingface.co/LiquidAI/LFM2-8B-A1B) model.
For more information about models, please read [the blog post](https://www.liquid.ai/company/news).

[HF PR](https://github.com/huggingface/transformers/pull/41401)
[GGUFs](https://huggingface.co/LiquidAI/LFM2-8B-A1B-GGUF)

* Do not use defaultdict

* Address PR feedback
This commit is contained in:
Tarek Dakhran
2025-10-07 20:03:35 +02:00
committed by GitHub
parent df1b612e29
commit aeaf8a36f0
7 changed files with 192 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
gguf-py/gguf/tensor_mapping.py
View File

@@ -358,6 +358,7 @@ class TensorNameMap:
"model.layers.{bid}.mlp.router", # openai-moe
"model.layers.{bid}.mlp.gate.wg", # hunyuan
"model.layers.{bid}.block_sparse_moe.primary_router", # smallthinker
"model.layers.{bid}.feed_forward.gate", # lfm2moe
),
MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
@@ -367,6 +368,7 @@ class TensorNameMap:
MODEL_TENSOR.FFN_EXP_PROBS_B: (
"model.layers.{bid}.mlp.gate.e_score_correction", # deepseek-v3 dots1
"model.layers.{bid}.mlp.moe_statics.e_score_correction", # ernie4.5-moe
"model.layers.{bid}.feed_forward.expert_bias", # lfm2moe
),
# Feed-forward up