Stream model reserve to Scaleway without local weight staging

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a09cbd-43c1-79f3-809e-1ee97b40b64d
This commit is contained in:
tegwick 2026-09-14 00:17:15 +02:00
parent ad88d16a52
commit 8a65297d56
16 changed files with 805 additions and 145 deletions

View file

@ -0,0 +1,25 @@
[Unit]
Description=Freedom Intelligence diskless model transfer to Scaleway
StartLimitIntervalSec=3600
StartLimitBurst=3
[Service]
Type=exec
WorkingDirectory=%h/.local/share/fi-reserve
Environment=PYTHONDONTWRITEBYTECODE=1
Environment=PYTHONUNBUFFERED=1
Environment=HF_HUB_DISABLE_TELEMETRY=1
Environment=FI_S3_CREDENTIAL_FILE=%t/fi-reserve/s3.json
ExecStart=/usr/bin/flock -n %t/fi-reserve/collector.lock %h/.local/share/fi-reserve/venv/bin/python %h/.local/share/fi-reserve/scripts/collect_model.py --repo-id deepseek-ai/DeepSeek-V4-Flash-0731 --local-name deepseek-ai__DeepSeek-V4-Flash-0731 --revision 7872f01b1d1fe23eabc4c98b48bffcef5a386062 --tier s --s3-bucket railiance-fi-open-weight-reserve
Restart=on-failure
RestartSec=60
TimeoutStopSec=210
MemoryHigh=384M
MemoryMax=512M
MemorySwapMax=0
CPUQuota=50%
TasksMax=32
UMask=0077
NoNewPrivileges=true
# Deliberately not enabled at boot: runtime credentials must be supplied first.