[Models] Add Phi3-mini, StableLM 1.6B, Qwen 1.8B, update MLC runtime #433
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
force-pushed
the
pr-0529-llama3-grammar
branch
from
1b946ed to
a4e18db
Compare
This was referenced May 29, 2024
CharlieFRuan
added a commit
that referenced
this pull request
May 30, 2024
### Changes Main changes include: - New prebuilt models: - Phi3-mini - StableLM-2-zephyr-1.6B - Qwen1.5-1.8B - Hermes2-Pro-Llama-3-8B to prebuilt models - Updates on `ModelRecord` fields - For detail see: #435 - Update all WASMs - For detail see: #433 - Update all WASMs to v0.2.39 - Support grammar for Llama3, hence update examples/json-mode to use `Llama3` and `Hermes2-pro-Llama3-8B` for function calling in `examples/json-schema` - Use `loglevel` package: - For details see #427 - Fix `index.js.map` issue for Vite - #420 - Enhance error handling and ServiceWorker ### TVMjs TVMjs compiled at apache/tvm@71f7af7 - Main changes include: - apache/tvm#17031 - apache/tvm#17028 - apache/tvm#17021 ### WASM version - All wasms updated to 0.2.39 via mlc-ai/binary-mlc-llm-libs#123 for new MLC-LLM runtime (mainly grammar)
jzhao62
pushed a commit
to jzhao62/web-llm
that referenced
this pull request
Dec 8, 2024
…lc-ai#433) This PR updates models to v0.2.39 compiled with mlc-ai/binary-mlc-llm-libs#123 The main change is the new MLC-LLM runtime, which supports grammar (i.e. json mode) for Llama3. - Hence we now read in field `tokenizer_info` (or deprecated `token_table_postproc_method`) from `mlc-chat-config.json` when post processing token table for Grammar - If neither is available, we use the default `byte_fallback` New prebuilt models introduced: - Phi3-mini-4k - Hermes-2-Pro-Llama-3-8B - Qwen1.5-1.8B - StableLM-2-zephyr_1.6B Updates on examples: - json-mode and json-schema now use Llama3 to demonstrate - Function calling inside json-schema now uses `Hermes-2-Pro-Llama-3-8B` instead of `Hermes-2-Pro-Mistral`
jzhao62
pushed a commit
to jzhao62/web-llm
that referenced
this pull request
Dec 8, 2024
### Changes Main changes include: - New prebuilt models: - Phi3-mini - StableLM-2-zephyr-1.6B - Qwen1.5-1.8B - Hermes2-Pro-Llama-3-8B to prebuilt models - Updates on `ModelRecord` fields - For detail see: mlc-ai#435 - Update all WASMs - For detail see: mlc-ai#433 - Update all WASMs to v0.2.39 - Support grammar for Llama3, hence update examples/json-mode to use `Llama3` and `Hermes2-pro-Llama3-8B` for function calling in `examples/json-schema` - Use `loglevel` package: - For details see mlc-ai#427 - Fix `index.js.map` issue for Vite - mlc-ai#420 - Enhance error handling and ServiceWorker ### TVMjs TVMjs compiled at apache/tvm@71f7af7 - Main changes include: - apache/tvm#17031 - apache/tvm#17028 - apache/tvm#17021 ### WASM version - All wasms updated to 0.2.39 via mlc-ai/binary-mlc-llm-libs#123 for new MLC-LLM runtime (mainly grammar)
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 updates models to v0.2.39 compiled with mlc-ai/binary-mlc-llm-libs#123
The main change is the new MLC-LLM runtime, which supports grammar (i.e. json mode) for Llama3.
tokenizer_info(or deprecatedtoken_table_postproc_method) frommlc-chat-config.jsonwhen post processing token table for Grammarbyte_fallbackNew prebuilt models introduced:
Updates on examples:
Hermes-2-Pro-Llama-3-8Binstead ofHermes-2-Pro-Mistral