nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 19 Packages Projects Releases Wiki Activity

mtmd: add MiMo v2.5 vision (#22883)

Browse Source 
* mimo-v2.5: vision support

* mimo-v2.5: use fused qkv for vision

* mimi-v2.5: fix f16 vision overflow

* mimo-v2.5: comment cleanups

* mimo-v2.5: Flash doesn't have mmproj
more cleanup
remember to use filter_tensors

* mimo-v2.5: fix trailing whitespace
This commit is contained in:
AesSedai
2026-05-12 11:11:14 +02:00
committed by GitHub
parent 78fbbc2c07
commit 4178259130
12 changed files with 460 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
gguf-py/gguf/tensor_mapping.py
+4
View File
@@ -1569,6 +1569,10 @@ class TensorNameMap:
"vision_model.transformer.resblocks.{bid}.attn.out_proj", # Step3-VL
),
MODEL_TENSOR.V_ENC_ATTN_SINKS: (
"visual.blocks.{bid}.attn.sinks", # mimovl
),
MODEL_TENSOR.V_ENC_POST_ATTN_NORM: (
"vision_tower.vision_model.encoder.layers.{bid}.layer_norm2",
"model.vision_tower.encoder.layers.{bid}.layer_norm2", # minicpmv4_6