2 days ago
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: draft model main-fp8
-
Run r1: draft model main-bf16
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Record memory, disk and GPU use with every log upload; use the built-in sampler
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Run r1: logs and results
-
Keep engine and compiler caches on the work disk
-
Engine-based data generation and an unattended GPU driver
4 days ago
-
Draft-model pipeline for the Speculative Decoding Championship
-
initial commit