arcee-ai mergekit issues

arcee-ai / mergekit

Tools for merging pretrained large language models.

GNU Lesser General Public License v3.0

3.97k stars 344 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Gemma2 support

#359 cg123 closed 5 hours ago
0
Support Qwen2 models with tied weights

#358 cg123 closed 6 hours ago
0
NuSLERP

#357 cg123 opened 3 days ago
0
Support of BitNet

#356 Ttimofeyka opened 3 days ago
0
About the merging method used for Arcee-Spark

#355 daiquocnguyen opened 4 days ago
0
Why all inputs to a slice must contain the same number of layers?

#354 Mihaiii closed 1 week ago
1
Sincerely.. I have no words...without insulting someone.

#353 0wwafa opened 1 week ago
2
MOE models that based on Gemma cannot work

#352 pharaohcaptain opened 1 week ago
1
New method: MAP:Low-compute Model Merging with Amortized Pareto Fronts via Quadratic Approximation

#351 duyvuleo opened 1 week ago
0
qwen2-0.5B cannot be merged into MoE

#350 letterk closed 6 hours ago
1
pass eval arguments directly in mergekit-evo

#349 johnwee1 opened 2 weeks ago
0
"Missing required parameter weight" for LoRA merge.

#348 bubbabug closed 2 weeks ago
2
Add out_dtype option

#347 cg123 closed 2 weeks ago
0
Fix DBZ Error

#346 MonsterAzi closed 1 week ago
0
New merge method / algorithm proposal: Geometric Median and TGMD merge

#345 6DammK9 opened 2 weeks ago
0
How to merge specific parameters

#344 qhz991029 opened 2 weeks ago
0
ValueError: operands could not be broadcast together with shapes (8192,28672) (8192,24576)

#343 ashmalvayani opened 2 weeks ago
1
Need some help in merging same architectures, but with different tokens in their tokenizers

#342 choprahetarth opened 3 weeks ago
4
Evolutionary Merging out of memory

#341 ArcherShirou opened 3 weeks ago
3
Weights Metrics

#340 ElliotStein opened 3 weeks ago
0
Use logscale for operations dealing with norm layers?

#339 jukofyork opened 3 weeks ago
0
parameters: int8_mask: true ??

#338 David-AU-github opened 4 weeks ago
1
Support for microsoft/Phi-3-vision-128k-instruct

#337 AshD opened 4 weeks ago
0
Questions about density gradient and weight gradient in Ties example

#336 zhang-tuo-pdf opened 4 weeks ago
0
Merge arbitrary pytorch models

#335 cg123 opened 4 weeks ago
0
Tokenizer merging overhaul

#334 cg123 opened 4 weeks ago
2
`extract_lora.py` improvements and fixes

#333 jukofyork opened 1 month ago
6
Add --load-in-4bit and --load-in-8bit for HF eval backend

#332 cg123 opened 1 month ago
0
Bump dependencies

#331 cg123 closed 1 month ago
0
Fix mergekit-moe for older python

#330 cg123 closed 1 month ago
0
Does this run single core?

#329 Autumnlight02 opened 1 month ago
2
`extract_lora.py` can't handle mismatched `lm_head` tensor due to added tokens

#328 jukofyork opened 1 month ago
10
How to merge only at q_proj layers with SLERP?

#327 yiyiwwang closed 1 month ago
2
Require later version of transformers

#326 tleyden closed 1 month ago
1
how to merge for different rope scaling?

#325 sangmandu opened 1 month ago
3
Merging models with different structures in linear

#324 HawkClaws opened 1 month ago
2
EvoMerge Genome Bug

#323 Jacobsolawetz closed 3 weeks ago
2
Relax dependency versions

#321 Nugine opened 1 month ago
0
Merge of hidden_size

#320 win10ogod opened 1 month ago
0
How to merge a VLM and LLM with different model type.

#319 tanyakansal30 opened 1 month ago
0
Add support for `subfolder` loading

#318 xzuyn opened 1 month ago
0
Mixed Precision Merging

#316 sais-github opened 1 month ago
1
Fix size check in TensorWriter.save_tensor

#315 cg123 closed 1 month ago
0
[mergekit-extract-lora] cast to float before running SVD

#314 cg123 closed 1 month ago
1
Fix Qwen MoE architecture check

#313 cg123 closed 1 month ago
0
Qwen/Qwen1.5-1.8B MoE Merging fails

#312 dgolchin closed 1 month ago
3
Existing Mergekit algorithms to merge VLM with LLM?

#310 ChaseKolozsy opened 1 month ago
1
Implementation of AdaMerging: Adaptive Model Merging for Multi-Task Learning

#309 varunlmxd opened 1 month ago
0
Merge only the transformer parts (including the input embedding layer)

#308 hank0316 opened 1 month ago
5
`mergekit-evolve` improvements

#307 cg123 closed 1 month ago
0