chore(neuron/beast): switch default-model quant from q5k to q6k
Some checks failed
CI / Format (push) Successful in 35s
build-prerelease / Resolve version stamps (push) Successful in 39s
CI / Clippy (push) Successful in 2m22s
build-prerelease / Build neuron-blackwell (push) Successful in 3m35s
CI / Test (push) Successful in 5m8s
build-prerelease / Build cortex binary (push) Successful in 4m34s
CI / Build cortex SRPM (push) Has been skipped
CI / Build neuron SRPM (push) Has been skipped
CI / Publish cortex to COPR (push) Has been skipped
CI / Publish neuron to COPR (push) Has been skipped
CI / Bump version in source (push) Has been skipped
build-prerelease / Package cortex RPM (push) Successful in 1m16s
build-prerelease / Build neuron-ampere (push) Successful in 5m12s
build-prerelease / Package helexa-neuron-ada RPM (push) Has been cancelled
build-prerelease / Package helexa-neuron-ampere RPM (push) Has been cancelled
build-prerelease / Package helexa-neuron-blackwell RPM (push) Has been cancelled
build-prerelease / Publish to rpm.lair.cafe (unstable) (push) Has been cancelled
build-prerelease / Build neuron-ada (push) Has been cancelled

q5k produced NaN logits on Qwen/Qwen3.6-27B under candle TP=2 (sampler
fell over with "logits unhealthy nan: 248320/248320"). q6k is the
quant that worked well in production under mistral.rs on the same
hardware, so it's the right baseline for verifying the mempool-trim
fix.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-27 12:36:18 +03:00
parent cdf0f4e66d
commit 740299bd9d

View File

@@ -19,6 +19,6 @@ name = "candle"
[[default_models]] [[default_models]]
model_id = "Qwen/Qwen3.6-27B" model_id = "Qwen/Qwen3.6-27B"
harness = "candle" harness = "candle"
quant = "q5k" quant = "q6k"
tensor_parallel = 2 tensor_parallel = 2
devices = [0, 1] devices = [0, 1]