Skip to content

tests : add test-jinja -py option for cross-checking#18906

Merged
ngxson merged 6 commits intoggml-org:masterfrom
ngxson:xsn/jinja-test-py
Jan 18, 2026
Merged

tests : add test-jinja -py option for cross-checking#18906
ngxson merged 6 commits intoggml-org:masterfrom
ngxson:xsn/jinja-test-py

Conversation

@ngxson
Copy link
Copy Markdown
Contributor

@ngxson ngxson commented Jan 17, 2026

Small QoL feature for test-jinja: Running test-jinja -py will output the results using the original jinja2 python. This is meant to be used to make sure that the test results are all aligned to the reference jinja2 implementation.

The way it works is by simply spawn python3 as a subprocess. Requires to have jinja2 pre-installed via pip

Some tests are currently failing on this mode, which I'm not 100% sure while. May worth investigating.

Note: I haven't tested this on windows

@ngxson ngxson requested review from CISC and aldehir January 17, 2026 23:14
@ngxson ngxson requested a review from ggerganov as a code owner January 17, 2026 23:14
@ngxson ngxson changed the title tests : add test-jinja -py option or cross-checking tests : add test-jinja -py option for cross-checking Jan 17, 2026
@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

Some tests are currently failing on this mode, which I'm not 100% sure while. May worth investigating.

Will take a look.

@github-actions github-actions Bot added the testing Everything test related label Jan 17, 2026
@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 17, 2026

Here are some of the failed tests for example. Looks quite strange. I've upgraded to python 3.14 and latest jinja2==3.1.6

  tojson ensure_ascii=true
    Python script failed with exit code: 1
    Output: Traceback (most recent call last):
  File "<string>", line 14, in <module>
(REDACTED)
    raise rewrite_traceback_stack(source=source)
  File "<template>", line 1, in top-level template code
TypeError: do_tojson() got an unexpected keyword argument 'ensure_ascii'

  ASSERTION FAILED : python execution
  tojson ensure_ascii=true (1 of 1 assertion(s) failed)                         [FAIL]
  tojson sort_keys=true
    Python script failed with exit code: 1
    Output: Traceback (most recent call last):
  File "<string>", line 14, in <module>
(REDACTED)
    raise rewrite_traceback_stack(source=source)
  File "<template>", line 1, in top-level template code
TypeError: do_tojson() got an unexpected keyword argument 'sort_keys'

  ASSERTION FAILED : python execution
  tojson separators=(',',':') (1 of 1 assertion(s) failed)                      [FAIL]
  tojson separators=(',',': ') indent=2
    Python script failed with exit code: 1
    Output: Traceback (most recent call last):
  File "<string>", line 14, in <module>
(REDACTED)
    raise rewrite_traceback_stack(source=source)
  File "<template>", line 1, in top-level template code
TypeError: do_tojson() got an unexpected keyword argument 'separators'

Comment thread tests/test-jinja.cpp
@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

Here are some of the failed tests for example. Looks quite strange. I've upgraded to python 3.14 and latest jinja2==3.1.6

transformers modify jinja2 a bit, see above.

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

Ah, sorry, a bit too quick on the cut'n'paste. :)

@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 17, 2026

no problem, I was thinking about the same possibility, just haven't dig through the correct place in transformers :)

one thing left though, seems like ImmutableSandboxedEnvironment cause in-place modifications like arr.append() to throw jinja2.exceptions.SecurityError

@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 17, 2026

Alright, just one problem left: jinja2.exceptions.UndefinedError: 'list object' has no attribute 'map'

@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

one thing left though, seems like ImmutableSandboxedEnvironment cause in-place modifications like arr.append() to throw jinja2.exceptions.SecurityError

And rightly so, this will happen with transformers too.

@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

You can verify transformers behavior using my Chat Template Editor BTW. :)

@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 17, 2026

I think I got it, map is a filter so it must be used as arr|map() and not arr.map()

However, it now introduce yet another set of errors:

  • jinja2.exceptions.TemplateRuntimeError: No filter named 'age'
  • jinja2.exceptions.TemplateRuntimeError: No filter named 0

@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

I think I got it, map is a filter so it must be used as arr|map() and not arr.map()

However, it now introduce yet another set of errors:

* `jinja2.exceptions.TemplateRuntimeError: No filter named 'age'`

* `jinja2.exceptions.TemplateRuntimeError: No filter named 0`

Yeah, without attribute map will apply the argument as a filter.

@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 17, 2026

should be fine now, I'm done with this. @CISC feel free to push commits directly here if you need to add something else

@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 17, 2026

I think I got it, map is a filter so it must be used as arr|map() and not arr.map()
However, it now introduce yet another set of errors:

* `jinja2.exceptions.TemplateRuntimeError: No filter named 'age'`

* `jinja2.exceptions.TemplateRuntimeError: No filter named 0`

Yeah, without attribute map will apply the argument as a filter.

Fixed that in #18883

Copy link
Copy Markdown
Member

@CISC CISC left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Super useful.

We should probably go back to ImmutableSandboxedEnvironment and disallow pop and append though, transformers uses this and I see no value in mutable arrays.

Copy link
Copy Markdown
Contributor

@aldehir aldehir left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I like this approach! I was worried about adding a dependency on Python esp. since it requires Jinja2, but the flag addresses that concern.

@ngxson ngxson merged commit fe44d35 into ggml-org:master Jan 18, 2026
75 of 78 checks passed
@ngxson
Copy link
Copy Markdown
Contributor Author

ngxson commented Jan 18, 2026

We should probably go back to ImmutableSandboxedEnvironment and disallow pop and append though, transformers uses this and I see no value in mutable arrays.

Actually it's a bit puzzling to me why the Kimi-K2-Instruct.jinja template inside models/templates use append() and pop(). It seems like a patch, as the original model doesn't have it.

@CISC
Copy link
Copy Markdown
Member

CISC commented Jan 18, 2026

We should probably go back to ImmutableSandboxedEnvironment and disallow pop and append though, transformers uses this and I see no value in mutable arrays.

Actually it's a bit puzzling to me why the Kimi-K2-Instruct.jinja template inside models/templates use append() and pop(). It seems like a patch, as the original model doesn't have it.

Ah, minja supports it, guess whoever added that only tested here.

Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
* tests : add test-jinja -py option or cross-checking

* Update tests/test-jinja.cpp

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

* fix + add source

* SandboxedEnvironment

* fix array.map case

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
phuongncn pushed a commit to phuongncn/llama.cpp-gx10-dgx-sparks-deepseekv4 that referenced this pull request Apr 28, 2026
… and new jinja template engine (ggml-org#1369)

---------

Co-authored-by: Piotr Wilkin <piotr.wilkin@syndatis.com>

common : add nemotron 3 parsing (ggml-org#18077)

common : add parser for ministral/mistral large 3/devstral 2 (ggml-org#17713)

common : default content to an empty string (ggml-org#18485)

chat: make tool description and parameters optional per OpenAI spec (ggml-org#18478)

Per the OpenAI API specification, both 'description' and 'parameters'
fields in tool function definitions are optional. Previously, the parser
would throw an exception if these fields were missing.

Attempts to fix ggml-org#17667

common : implement new jinja template engine (ggml-org#18462)
---------

Co-authored-by: Alde Rojas <hello@alde.dev>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

jinja: correct member access rule (ggml-org#18905)

jinja : fix lexing of float literals with sign (ggml-org#18901)

jinja : add missing tojson filter for bool (ggml-org#18900)

jinja : attribute support for join, map and sort (ggml-org#18883)

jinja : fix object item order (and properly implement dictsort) (ggml-org#18904)

tests : add test-jinja -py option for cross-checking (ggml-org#18906)

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

ci : run test-jinja -py on high perf [no ci] (ggml-org#18916)

jinja : fix undefined keys and attributes and int/float as bool (ggml-org#18924)

jinja: support none|string (ggml-org#18995)

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

jinja : implement mixed type object keys (ggml-org#18955)

---------

Co-authored-by: Xuan Son Nguyen <son@huggingface.co>

jinja : undefined should be treated as sequence/iterable (return string/array) by filters/tests (ggml-org#19147)

`tojson` is not a supported `undefined` filter

keep it DRY and fix some types

jinja : do not pass empty tools and add some none filters (ggml-org#19176)

jinja : add unordered_map include to value.h [no ci] (ggml-org#19205)

jinja : add missing 'in' test to template engine (ggml-org#19004) (ggml-org#19239)

The jinja template parser was missing the 'in' test from
global_builtins(), causing templates using reject("in", ...),
select("in", ...), or 'x is in(y)' to fail with
"selectattr: unknown test 'in'".

This broke tool-calling for Qwen3-Coder and any other model
whose chat template uses the 'in' test.

Added test_is_in supporting array, string, and object containment
checks, mirroring the existing 'in' operator logic in runtime.cpp.

Includes test cases for all three containment types plus
reject/select filter usage.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

---------

Co-authored-by: Sid Mohan <sidmohan0@users.noreply.github.com>
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: Xuan Son Nguyen <son@huggingface.co>

Add Jinja support for "indent" string filter (ggml-org#19529)

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

---------

Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>

add vendor

refactor chat

server : support preserving reasoning_content in assistant message (ggml-org#18994)

chat : fix translategemma crash on common_chat_format_example (ggml-org#19019)

chat: fix language input for translategemma (ggml-org#19052)

Co-authored-by: Aldehir Rojas <hello@alde.dev>

---------

Co-authored-by: Aldehir Rojas <hello@alde.dev>

chat: fix case where template accepts type content only (ggml-org#19419)

mtmd : chat : Fix extra \n between text and media marker (ggml-org#19595)

Thanks to @tugot17 for detecting and reporting the issue.

For vision models (e.g. LFM2.5-VL-1.6B and Qwen/Qwen3-VL-4B-Instruct) `llama-mtmd-cli` produces identical output to HF implementation.

However `llama-server` doesn't. I traced it down to extra newline
inserted after `<__media__>`.

This happens in `to_json_oaicompat`, that treats media markers as text
and joins all parts with `\n` separator.

PR introduces new type `media_marker` and uses it for media markers.
Extra logic is added to prevent insertion of newlines before and after
media markers.

With this change number of input tokens is identical to HF
implementation and as a result the output is also identical.

I explored other ways to address the issue
* remove completely `\n` between text parts in `to_json_oaicompat`
* merge text messages in server-common.cpp before sending them to `to_json_oaicompat`

Please propose alternative ways of fixing this issue.

Co-authored-by: Piotr Wilkin (ilintar) <piotr.wilkin@syndatis.com>

---------

Co-authored-by: Piotr Wilkin (ilintar) <piotr.wilkin@syndatis.com>

common : merge qwen3-coder and nemotron nano 3 parsers (ggml-org#19765)

common : fix improper trimming in XML parser on complete message (ggml-org#19805)

Co-authored-by: Jules LEIDELINGER <11395311+julio75012@users.noreply.github.com>

jinja: correct stats for tojson and string filters (ggml-org#19785)

jinja : correct default size for string slices (ggml-org#19913)

common : handle unicode during partial json parsing (ggml-org#16526)

common : fix json schema with '\' in literals (ggml-org#17307)

add back qwen_coder_xml and mirothinker

Co-authored-by: Aldehir Rojas <hello@alde.dev>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

testing Everything test related

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants