unslothai/unsloth · warning · RuntimeError
load timed out
Error message
load timed out
What it means
Thrown by parseMaxOutputTokens (chat-providers-dialog.tsx:503-507) when the parsed integer is below CUSTOM_MAX_OUTPUT_TOKENS_MIN, defined as 64 in external-providers.ts:265. A provider-level output-token ceiling below 64 would truncate essentially every reply, so the dialog enforces a floor.
Source
Thrown at scripts/perf_verify.py:82
token = os.environ.get("HF_TOKEN")
def load(mode_speed = None, mode_mem = None):
backend.begin_load(
args.model,
gguf_filename = args.gguf,
hf_token = token,
speed_mode = mode_speed,
memory_mode = mode_mem,
)
deadline = time.time() + 2400
while time.time() < deadline:
ph = backend.load_progress().get("phase")
if ph == "ready":
return backend.status()
if ph == "error":
raise RuntimeError(f"load error: {backend.load_progress()}")
time.sleep(0.5)
raise RuntimeError("load timed out")
def gen():
torch.cuda.synchronize()
t0 = time.time()
img = backend.generate(
prompt = args.prompt,
width = args.width,
height = args.height,
steps = args.steps,
guidance = 0.0,
seed = args.seed,
batch_size = 1,
)["images"][0]
torch.cuda.synchronize()
return img, time.time() - t0
def timed(
mode_speed,View on GitHub (pinned to 203007d190)
Solutions
- Enter 64 or higher, e.g. '256' or '1024'.
- Leave the field empty to use the provider's default rather than trying to express 'unlimited' with 0.
Example fix
// before 16 // after 64
Defensive patterns
Strategy: validation
Validate before calling
const CUSTOM_MAX_OUTPUT_TOKENS_MIN = 64;
function meetsMinTokens(input: string): boolean {
const t = input.trim();
return t === '' || (/^\d+$/.test(t) && Number(t) >= CUSTOM_MAX_OUTPUT_TOKENS_MIN);
} Prevention
- Show 'Minimum 64' helper text under the field.
- Clamp small values up to the minimum on blur instead of erroring on submit.
- Document that 'no limit' is expressed by leaving the field empty, never by 0.
When it happens
Trigger: Entering a value from 0 to 63 (the regex allows leading zeros, so '0'..'0063' also hit it) into the Max Tokens limit field and saving the provider.
Common situations: Users entering '1' to 'minimize' answers; confusing max tokens with temperature-style small decimals; testing edge values; entering '0' expecting 'no limit'.
Understand the failure class
- Timeouts: ETIMEDOUT, deadlines, and hung requests — what actually expires when a request times out.
Related errors
- Cell uses shell metacharacters / interpolation but --no-allo
- load error: {backend.load_progress()}
- model load did not reach ready
- unknown family '{name}'
- Refused notebook fetch from {host!r}: not in allowlist {sort
AI-assisted analysis of unslothai/unsloth@203007d190 (2026-08-15).
Data as JSON: /api/errors/6daaa4e69e18c703.
Report an issue: GitHub.