Slow Upload Fix Recommendations | Generated by AI

Home 2026.06

Here’s the smoke test results:

config.json (831B)  →  1.7s  ✓
tokenizer.json (3.4MB) → 17.9s  ≈ 190 KB/s

190 KB/s throughput from your current connection. At that rate:

model.safetensors (475MB) → estimated ~42 minutes

And it will likely stall mid-transfer (like you saw with GCS).

Recommendations (ranked):

  1. hf-mirror.com — China-optimized HuggingFace mirror. Try:

    HF_ENDPOINT=https://hf-mirror.com hf upload lzwjava/sec-edgar-gpt-124m model.safetensors model.safetensors

  2. DMIT proxy (LA) — route through your hysteria2 proxy. The hf CLI respects https_proxy:

    https_proxy=http://127.0.0.1:<hysteria-port> hf upload lzwjava/sec-edgar-gpt-124m model.safetensors model.safetensors

  3. AMD Dev Cloud relay — rsync to Atlanta, upload from there (US→HF is fast)

Want me to try option 1 (hf-mirror) or option 2 (DMIT proxy) as the actual upload? Or just kick off the 475MB upload on current connection and hope it doesn’t stall?

Back Donate