[Grammar] Integrate with XGrammar #635
Merged
Conversation
This file contains 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
CharlieFRuan
added a commit
that referenced
this pull request
Nov 22, 2024
### Change - #635 - Integrate with `web-xgrammar` - Support `ResponseFormat.type == "grammar"`, where you specify an EBNF grammar string - Add `grammar_init_ms` and `grammar_per_token_ms` to `CompletionUsage.extra` when using grammar - Add `time_to_first_token_s` (TTFT) and `time_per_output_token_s` (TPOT), `e2e_latency_s` to `CompletionUsage.extra` - Add `ignore_eos` to `Completion` and `ChatCompletion` requests - #632 - Fixes of vram requirement for Qwen2.5-Coder-1.5B-Instruct model ### TVMjs - No change, version `0.18.0-dev2` just like 0.2.71
jzhao62
pushed a commit
to jzhao62/web-llm
that referenced
this pull request
Dec 8, 2024
This PR integrates with XGrammar: https://github.com/mlc-ai/xgrammar. Prior to this PR, grammar is supported by the grammar portion of MLC-LLM compiled into the model WASM. That portion is now a standalone project XGrammar. Therefore, this PR adds `mlc-ai/web-xgrammar` as part of the dependency and remove `src/grammar.ts`. We update `llm_chat.ts` accordingly for xgrammar's APIs. In addition, besides `json_schema`, we now also support requests with EBNF-formatted strings by using the following in the chat completion request. See `examples/json-schema`'s `ebnfGrammarExample()` for a full example. ```typescript response_format: { type: "grammar", grammar: jsonGrammarStr, } as webllm.ResponseFormat, ``` We also add the following performance info: - Add `grammar_init_ms` and `grammar_per_token_ms` to `CompletionUsage.extra` when using grammar - Add `time_to_first_token_s` (TTFT) and `time_per_output_token_s` (TPOT), `e2e_latency_s` to `CompletionUsage.extra` We also add `ignore_eos` to `Completion` and `ChatCompletion` requests, which can be useful for benchmarking purposes.
jzhao62
pushed a commit
to jzhao62/web-llm
that referenced
this pull request
Dec 8, 2024
### Change - mlc-ai#635 - Integrate with `web-xgrammar` - Support `ResponseFormat.type == "grammar"`, where you specify an EBNF grammar string - Add `grammar_init_ms` and `grammar_per_token_ms` to `CompletionUsage.extra` when using grammar - Add `time_to_first_token_s` (TTFT) and `time_per_output_token_s` (TPOT), `e2e_latency_s` to `CompletionUsage.extra` - Add `ignore_eos` to `Completion` and `ChatCompletion` requests - mlc-ai#632 - Fixes of vram requirement for Qwen2.5-Coder-1.5B-Instruct model ### TVMjs - No change, version `0.18.0-dev2` just like 0.2.71
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This PR integrates with XGrammar: https://github.com/mlc-ai/xgrammar.
Prior to this PR, grammar is supported by the grammar portion of MLC-LLM compiled into the model WASM. That portion is now a standalone project XGrammar. Therefore, this PR adds
mlc-ai/web-xgrammaras part of the dependency and removesrc/grammar.ts. We updatellm_chat.tsaccordingly for xgrammar's APIs.In addition, besides
json_schema, we now also support requests with EBNF-formatted strings by using the following in the chat completion request. Seeexamples/json-schema'sebnfGrammarExample()for a full example.We also add the following performance info:
grammar_init_msandgrammar_per_token_mstoCompletionUsage.extrawhen using grammartime_to_first_token_s(TTFT) andtime_per_output_token_s(TPOT),e2e_latency_stoCompletionUsage.extraWe also add
ignore_eostoCompletionandChatCompletionrequests, which can be useful for benchmarking purposes.