Skip to content

[P0] Threshold + Params for big model. Part 2. #361

Closed @tcharchian opened 2025-09-16 22:31 UTC 2 comments Updated 2025-12-05 22:18 UTC
  1. Threshold + Params for big model;
  2. Scripts to compute, re-check for existing;

  3. New Models

  4. gpt-oss-120b

  5. DeepSeek-R1-0528
  6. gemma-3-27b-it
  7. Qwen3-30B-A3B-Instruct-2507
  8. gpt-oss-20b

  9. Qwen3-235B

  10. Instruction to do it

  11. Inference Validation finetuning;
  12. Fine-tuning Qwen 235

💬 Comments (2)

@tcharchian commented 2025-12-05 22:16 UTC

GPT-OSS can be implemented after the vLLM update. Right now, it is being handled by community contributors from the bounty program https://discord.com/channels/1336477374442770503/1425189436748206171/1446142256900997152

@tcharchian commented 2025-12-05 22:18 UTC

The threshold-calculation task is completed for the models listed above (except GTP-OSS). They haven’t deployed it to the chain yet. They will most likely be deployed after the vLLM update

🔄 Auto-synced from Issue #361 every hour.