Repository navigation
Conversation
…ize the slots enforce max_tokens_per_sequence read llama_n_ctx_seq while every slot caps that at the model's training context. On the M5 (Qwen3.6-35B-A3B, -c 524285 --parallel 5 --kv-unified) the report said 524288 and each slot refused past 262144. The capped value is now one member, set at load, that sizes the slots and is reported.
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Why: the core's admission reads
context_geometry.max_tokens_per_sequenceas the most one mind's sequence may hold (our fork is the contract). It reportedllama_n_ctx_seqwhile every slot caps that at the model's training context (server-context.cpp, slot init).Measured (the M5, 2026-11-11 16:2xZ, Qwen3.6-35B-A3B,
-c 524285 --parallel 5 --kv-unified):/propscontext_geometry = {allocated_tokens 524288, max_tokens_per_sequence 524288, sequence_capacity 5},/slotsn_ctx = 262144on every slot. A grant above 262,144 would be refused by the slot the report said could hold it.Change: the capped value is one member (
n_ctx_slot), set at load; the slots are sized from it and the report reads it. No behaviour change for the slots.Validation: compiles in CI; live receipt after the next engine build:
/propsmax_tokens_per_sequence==/slotsn_ctx. Core side: CambrianTech/continuum#5947 (an admitted request is bounded by its grant alone).🤖 Generated with Claude Code
https://claude.ai/code/session_01LoTjvf5j3Ez13g6k8mRkFo