add chat template #42
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
WeiLai5432
force-pushed
the
feat/add_chat_template
branch
from
5ae9f2f to
60c7c38
Compare
WeiLai5432
force-pushed
the
feat/add_chat_template
branch
from
60c7c38 to
a360e42
Compare
daulet
reviewed
Aug 1, 2025
daulet
reviewed
Aug 1, 2025
daulet
reviewed
Aug 1, 2025
tokenizer.go
Outdated
| // defer C.free(unsafe.Pointer(cErr)) | ||
|
|
||
| cChatTemplate := C.new_chat_template(cTemplate, cBosToken, cEosToken) | ||
| // defer C.free_chat_template(cChatTemplate) |
Author
There was a problem hiding this comment.
I added a Close() method so users can free the resource when needed. We shouldn’t free cChatTemplate here as it’s returned and may be used afterwards.
daulet
reviewed
Aug 1, 2025
release/Dockerfile
Outdated
| # syntax=docker/dockerfile:1.4 | ||
|
|
||
| FROM rust:1.87 as builder-rust | ||
| FROM swr.cn-north-4.myhuaweicloud.com/ddn-k8s/docker.io/rust:1.87 as builder-rust |
Author
There was a problem hiding this comment.
Sorry, this should not be pushed.
daulet
reviewed
Aug 1, 2025
daulet
reviewed
Aug 1, 2025
daulet
reviewed
Aug 6, 2025
| [dependencies] | ||
| libc = "0.2.162" | ||
| tokenizers = {version = "0.20.2" } | ||
| tokenizers = { version = "0.20.0", features = ["http"] } |
Owner
There was a problem hiding this comment.
is there a reason for the downgrade in version? is something broken in 20.2?
daulet
reviewed
Aug 6, 2025
| // Construct the model URL | ||
| modelURL := fmt.Sprintf("%s/%s/resolve/main", baseURL, modelID) | ||
| modelURL := "" | ||
| hfEndpoint, exist := os.LookupEnv("HF_ENDPOINT") |
Owner
There was a problem hiding this comment.
if we gonna make this public config, lets spell it out HUGGINGFACE_ENDPOINT
daulet
reviewed
Aug 6, 2025
| if err != nil { | ||
| t.Fatalf("Failed to apply chat template: %v", err) | ||
| } | ||
| fmt.Println(result) |
Owner
There was a problem hiding this comment.
lets assert expected output, instead of printing it
daulet
reviewed
Aug 6, 2025
Comment on lines
+197
to
+211
| // TODO: replace with better solution | ||
| // hack to adjust gemma3 template for debug | ||
| // replace 'messages[0]['content'][0]['text']' with 'messages[0]['content']' | ||
| let mutated_template = template.replace( | ||
| "messages[0]['content'][0]['text']", | ||
| "messages[0]['content']", | ||
| ); | ||
| // Hack to fix Qwen3 templating. | ||
| // It uses python notation to reverse lists, which do not exist in minijinja | ||
| // so we're using the reverse filter instead. | ||
| let mutated_template = mutated_template.replace("[::-1]", "|reverse"); | ||
| // TODO: replace with a better solution | ||
| // Hack to remove the {% generation %} and {% endgeneration %} statements from | ||
| // the Jinja2 chat templates if there, since those are only using for assistant | ||
| // masking during training, and should be ignored during inference |
Owner
There was a problem hiding this comment.
should these hacks live in higher layer, like in go, so the library doesnt have to inherit them? Or maybe some kind of config preprocessing script to correct them?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.