evan/nix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(llama-swap): add --default-chat-template-kwargs to vLLM 3090 configs

Browse Source 
Sync all three Qwen3.6 27B vLLM configs (tools-text, long-text,
long-vision) with club-3090 83bf73d. Adds disable-thinking flag
and introduces upstream hash tracking comments for future syncs.

Update update-vllm-3090-configs skill to use hash-based skip logic.
This commit is contained in:
Evan Reichard
2026-05-10 16:57:17 -04:00
parent 885fe8517d
commit 6fff658f9d
2 changed files with 23 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
.agents/skills/update-vllm-3090-configs/SKILL.md
View File

@@ -19,6 +19,15 @@ Local config keys:
- `vllm-qwen3.6-27b-long-text`
- `vllm-qwen3.6-27b-long-vision`
## Hash Tracking
Each config entry stores an upstream commit hash comment:
`# Upstream: club-3090 <hash> (<date>) - <compose-file>`
When comparing, first extract stored hashes. If a config's hash matches
upstream HEAD, skip it (report "already synced"). Only full-diff configs
whose hash differs. Update the hash comment when edits are applied.
## Upstream References
Compare against `club-3090` master:
@@ -37,25 +46,26 @@ git clone https://github.com/noonghunna/club-3090 _scratch/club-3090 2>/dev/null
## Required Workflow
1. Fetch/update upstream refs under `_scratch/club-3090` or fetch the raw files.
2. Compare upstream compose files to the three local llama-swap entries. Translate docker-compose semantics into the existing `docker run`/llama-swap format.
3. Compare upstream `scripts/setup.sh` Genesis pin to local `GENESIS_PIN` in `setup-qwen36-vllm.sh`.
4. Check upstream compose volumes/entrypoint for sidecar patches. If patches are added, removed, renamed, or invoked differently, update both:
2. Extract stored upstream hashes from `# Upstream: club-3090 ...` comments in config.nix. Skip any config whose hash matches upstream HEAD (report "already synced").
3. Compare upstream compose files to the remaining local llama-swap entries. Translate docker-compose semantics into the existing `docker run`/llama-swap format.
4. Compare upstream `scripts/setup.sh` Genesis pin to local `GENESIS_PIN` in `setup-qwen36-vllm.sh`.
5. Check upstream compose volumes/entrypoint for sidecar patches. If patches are added, removed, renamed, or invoked differently, update both:
- runtime mounts and `python3 /patches/...` calls in `config.nix`
- download/install logic and summary in `setup-qwen36-vllm.sh`
5. Ignore these diffs unless the user explicitly asks otherwise:
6. Ignore these diffs unless the user explicitly asks otherwise:
- `shm_size` / shm-related compose settings
- local timing patch `patch_timings_07351e088.py` and its mount/invocation
- model served-name differences caused by llama-swap `${MODEL_ID}`
- `HUGGING_FACE_HUB_TOKEN`; keep local CUDA device/env choices
- upstream relative paths vs local `/mnt/ssd/vLLM/...` paths
- docker-compose format vs local llama-swap/Nix format
6. Before editing, present:
7. Before editing, present:
- upstream files/commit checked
- meaningful diffs found
- ignored diffs
- exact planned local changes
Then wait for explicit user approval.
7. After approval, edit minimally and validate:
8. After approval, edit minimally and update the `# Upstream: club-3090 ...` hash comments. Validate:
- `bash -n modules/nixos/services/llama-swap/setup-qwen36-vllm.sh`
- `nix-instantiate --parse modules/nixos/services/llama-swap/config.nix`
8. Summarize changed files and any remaining upstream differences.
9. Summarize changed files and any remaining upstream differences.