Skip to content

feat: add LFM2.5-1.2B-Instruct-DFlash training recipe#556

Open
nathanrchn wants to merge 1 commit into
sgl-project:mainfrom
nathanrchn:lfm
Open

feat: add LFM2.5-1.2B-Instruct-DFlash training recipe#556
nathanrchn wants to merge 1 commit into
sgl-project:mainfrom
nathanrchn:lfm

Conversation

@nathanrchn
Copy link
Copy Markdown

  • New draft config configs/lfm2.5-1.2b-instruct-dflash.json (8-layer Qwen3-style draft, vocab 65536, block_size 16).
  • New torchrun launcher examples/run_lfm2.5_1.2b_instruct_dflash_online.sh targeting LiquidAI/LFM2.5-1.2B-Instruct via the sglang backend.
  • New "lfm" chat template (Qwen-style tokens, empty system prompt) so the launcher can pass --chat-template lfm without altering the existing "qwen" template.

- New draft config configs/lfm2.5-1.2b-instruct-dflash.json (8-layer Qwen3-style draft, vocab 65536, block_size 16).
- New torchrun launcher examples/run_lfm2.5_1.2b_instruct_dflash_online.sh targeting LiquidAI/LFM2.5-1.2B-Instruct via the sglang backend.
- New "lfm" chat template (Qwen-style tokens, empty system prompt) so the launcher can pass --chat-template lfm without altering the existing "qwen" template.
@gemini-code-assist
Copy link
Copy Markdown
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant