books/gu-zhenren/rerun2/pipeline-arm-mistral-rerun2.yaml.bak-pre-pack15

73 lines
1.9 KiB
Text

# Pere-run rerun2 pipeline — SWAP ARM: editor mistral-large-2512 (WS6/D39.9). Identical to
# pipeline-rerun2.yaml EXCEPT edit.model glm-5 -> mistral-large-2512. Different editor-model =>
# different stageSnap.Model => different snapshot_edit (separate run, doesn't conflict with base).
# few_shot default ON (mistral is NOT a reasoning model; P1a discourse few-shot doesn't disturb its CoT).
# ⚠ Rate-guard: models.yaml carries rate_limit{max_concurrency:2} on mistral-large-2512; the wave
# executor's per-model semaphore caps concurrency (~48% retry-fails without it). No config change needed here.
core: C1
version: 1
defaults:
max_output_ratio: 2.2
min_max_tokens: 2048
context:
glossary_injection: selective
glossary_token_budget: 800
cache_ttl: "5m"
segmentation:
draft_budget_out: 1797
edit_ceiling_out: 3200
fertility:
cjk: 1.1978
other: 0.3852
waves:
workers: 4
retries:
regenerate_before_escalate: 1
stages:
- name: draft
role: translator
model: deepseek-v4-flash
prompts:
zh-ru: /home/ubuntu/projects/textmachine/backend/prompts/translator.md
prompt_version: v1-reflow
temperature: 0.3
reasoning: "off"
escalate_to: deepseek-v4-pro
- name: edit
role: editor
model: mistral-large-2512
prompts:
zh-ru: /home/ubuntu/projects/textmachine/backend/prompts/editor.md
prompt_version: v3-discourse-reflow
temperature: 0.4
reasoning: "off"
gates:
coverage:
enabled: false
len_ratio_bounds:
zh-ru: [2.2, 4.2]
ja-ru: [1.4, 2.6]
en-ru: [0.70, 1.4]
sent_cov_min: 0.75
min_chunk_chars: 500
sanitizer:
enabled: true
escalation:
chains:
default: [deepseek-v4-pro, glm-5.1, gemini-3.1-pro-preview]
adult: [grok-4.3]
budget_usd: 3.0
fanout:
candidates: 1
mining:
contrast_path: /home/ubuntu/projects/textmachine/eval/exp16/data/jieba_dict_general_zh.txt