Open Jasmin68k opened 2 weeks ago
Thanks for trying aider and filing this issue.
You can set the edit mode with --edit-format
.
Thank you!
The reason I opened the issue is more about
Suggesting to add vertex_ai/gemini-pro-experimental
to the known models with the same edit mode as gemini/gemini-1.5-pro-exp-0827
, since it's the same model, assuming that the reason for Aider to chose whole
edit mode for this model is that it does not know of the model.
Wondering what actually is the reason for Aider chosing diff
as opposed to diff-fenced
for gemini/gemini-1.5-pro-002
, while diff-fenced
is being described to be the proper edit mode in https://aider.chat/docs/leaderboards/ (sorry, had originally added wrong link in op for this) and making sure the proper/best edit mode is automatically chosen and the information in the leaderboard is correct.
Done.
The change is available in the main branch. You can get it by installing the latest version from github:
aider --install-main-branch
# or...
python -m pip install --upgrade --upgrade-strategy only-if-needed git+https://github.com/Aider-AI/aider.git
If you have a chance to try it, let me know if it works better for you.
Thanks!
I tested release v0.63.0:
diff-fenced
for vertex_ai/gemini-pro-experimental
now, so that is fixed.gemini-1.5-pro-002
to use diff-fenced
, unless there's a specific reason you chose diff
for this version of Gemini Pro, then the leaderboard needs updating.
Issue
Hello!
Just two observations with v0.60.1:
aider uses
whole
edit mode forvertex_ai/gemini-pro-experimental
(probably because it doesn't know it), but the model is the same asgemini/gemini-1.5-pro-exp-0827
, see https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/gemini-experimental, so it should probably usediff-fenced
, too.aider uses
diff
edit mode forgemini/gemini-1.5-pro-002
, while all other Gemini Pro models usediff-fenced
. Onhttps://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/gemini-experimentalhttps://aider.chat/docs/leaderboards/, it saysdiff-fenced
also forgemini/gemini-1.5-pro-002
. I don't know, which mode is better for the model (maybe the newer Gemini Pro really works better with diff?), but either aider's model config or the benchmark page is wrong.Hope it helps.
Version and model info
Aider v0.60.1 Model: gemini/gemini-1.5-pro-exp-0827 // vertex_ai/gemini-pro-experimental // gemini/gemini-1.5-pro-002