TBNilles/qwen3.5-122B-A10B-on-spark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: fix Reproducing — --start defaults to dense now; dflash is the explicit opt-in

Browse Source
This commit is contained in:
ent
2026-06-24 16:46:22 +10:00
parent 4e5c9552fe
commit 1d0f2e1af3
1 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+5 -5
View File
@@ -317,13 +317,13 @@ docs/
## Reproducing ## Reproducing
```bash ```bash
# default agent path (DFlash) + smoke test: # default: the dense profile (downloads the prebuilt hybrid) + smoke test:
./install.sh --start ./install.sh --start
# the dense stack (downloads the prebuilt hybrid checkpoint):
./install.sh --start --profile dense
# (to build the hybrid checkpoint locally instead of downloading it:) # (to build the hybrid checkpoint locally instead of downloading it:)
# ./install.sh --build-hybrid && ./install.sh --start --profile dense # ./install.sh --build-hybrid && ./install.sh --start
# plain-DFlash agent path (serves the Intel INT4 checkpoint, largest KV pool):
./install.sh --start --profile dflash
# benchmarks (run on the host against the server; requires: pip install requests): # benchmarks (run on the host against the server; requires: pip install requests):
python3 scripts/bench_decode.py --base-url http://127.0.0.1:8000 --model qwen \ python3 scripts/bench_decode.py --base-url http://127.0.0.1:8000 --model qwen \