Conversation
| prithivMLmods/Qwen2.5-7B-DeepSeek-R1-1M | ||
| prithivMLmods/QwQ-Math-IO-500M | ||
| prithivMLmods/Triangulum-v2-10B | ||
| qingy2024/Falcon3-2x10B-MoE-Instruct |
There was a problem hiding this comment.
Is this removal accidental?
There was a problem hiding this comment.
This model is removed from HF
https://huggingface.co/qingy2024/Falcon3-2x10B-MoE-Instruct
| EXPECT_THAT([]() { render("{% if 1 %}{% else %}{% elif 1 %}{% endif %}", {}, {}); }, ThrowsWithSubstr("Unterminated if")); | ||
| EXPECT_THAT([]() { render("{% filter trim %}", {}, {}); }, ThrowsWithSubstr("Unterminated filter")); | ||
| EXPECT_THAT([]() { render("{# ", {}, {}); }, ThrowsWithSubstr("Missing end of comment tag")); | ||
|
|
There was a problem hiding this comment.
Could you please also add an unterminated call test:
EXPECT_THAT([]() { render("{%- call test -%}", {}, {}); }, ThrowsWithSubstr("Missing end of call tag"));There was a problem hiding this comment.
Tests for call and macro added
ochafik
left a comment
ochafik
left a comment
There was a problem hiding this comment.
Thanks again @yatarkan !
I've just lost write access to this repo (left Google over 3 months ago 🤪), @matthewchan-g @Ferev could one of you please merge this one?
|
@ochafik I'm happy to help with merging the PR. There are currently 8 failing checks. Should the PR still be merged or should these failing checks be fixed first? |
|
It looks like the tests are failing because they are attempting to get gated files from HF. |
|
Seems this introduces a memory leak: |
4 participants
Some models (e.g.
openbmb/MiniCPM3-4B) have macro in their chat templates and execute them with call blocks.This PR:
call/endcallfor calling macro (w/ macro arguments andcaller()syntax) and passing a macro to another macroopenbmb/MiniCPM3-4Bto test modelsFixes #73