claw/diarization-ui
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(diarization-ui): raise default num_predict to 16384

Browse Source 
Thinking tokens count against num_predict. At 4096 the model was
running out mid-response after spending ~3000 tokens on thinking.
16384 gives enough headroom for thinking + full response.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Wolf G. Beckmann
2026-05-06 16:26:55 +02:00
parent 831a29a23a
commit 39250e6582
2 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.env.example
View File

@@ -1,5 +1,5 @@
API_BASE=http://gx10.aquantico.lan:8093 API_BASE=http://gx10.aquantico.lan:8093
OLLAMA_BASE_URL=http://gx10.aquantico.lan:11434 OLLAMA_BASE_URL=http://gx10.aquantico.lan:11434
OLLAMA_MODEL=qwen3.5:9b OLLAMA_MODEL=qwen3.5:9b
OLLAMA_NUM_PREDICT=4096 OLLAMA_NUM_PREDICT=16384
OLLAMA_THINK=true OLLAMA_THINK=true

2
app.py
View File

@@ -15,7 +15,7 @@ from fastapi.responses import HTMLResponse, PlainTextResponse, Response, JSONRes
API_BASE = os.getenv("API_BASE", "http://gx10.aquantico.lan:8093").rstrip("/") API_BASE = os.getenv("API_BASE", "http://gx10.aquantico.lan:8093").rstrip("/")
OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://gx10.aquantico.lan:11434").rstrip("/") OLLAMA_BASE_URL = os.getenv("OLLAMA_BASE_URL", "http://gx10.aquantico.lan:11434").rstrip("/")
OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "qwen3.5:9b") OLLAMA_MODEL = os.getenv("OLLAMA_MODEL", "qwen3.5:9b")
OLLAMA_NUM_PREDICT = int(os.getenv("OLLAMA_NUM_PREDICT", "4096")) OLLAMA_NUM_PREDICT = int(os.getenv("OLLAMA_NUM_PREDICT", "16384"))
OLLAMA_THINK = os.getenv("OLLAMA_THINK", "true").lower() in ("1", "true", "yes") OLLAMA_THINK = os.getenv("OLLAMA_THINK", "true").lower() in ("1", "true", "yes")
DB_PATH = os.getenv("DB_PATH", "/data/ui.db") DB_PATH = os.getenv("DB_PATH", "/data/ui.db")