Image-Text-to-Text
MLX
Safetensors
English
inkling2_loop
mlx-vlm
inkling2
Mixture of Experts
quantized
gptq
apple-silicon
tool-calling
agents
conversational
8-bit precision
Instructions to use VisualInference/Inkling-2-Mini-MLX-2bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use VisualInference/Inkling-2-Mini-MLX-2bit with MLX:
# Make sure mlx-vlm is installed # pip install --upgrade mlx-vlm from mlx_vlm import load, generate from mlx_vlm.prompt_utils import apply_chat_template from mlx_vlm.utils import load_config # Load the model model, processor = load("VisualInference/Inkling-2-Mini-MLX-2bit") config = load_config("VisualInference/Inkling-2-Mini-MLX-2bit") # Prepare input image = ["http://images.cocodataset.org/val2017/000000039769.jpg"] prompt = "Describe this image." # Apply chat template formatted_prompt = apply_chat_template( processor, config, prompt, num_images=1 ) # Generate output output = generate(model, processor, formatted_prompt, image) print(output) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Pi
How to use VisualInference/Inkling-2-Mini-MLX-2bit with Pi:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "VisualInference/Inkling-2-Mini-MLX-2bit"
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "mlx-lm": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "VisualInference/Inkling-2-Mini-MLX-2bit" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Hermes Agent
How to use VisualInference/Inkling-2-Mini-MLX-2bit with Hermes Agent:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "VisualInference/Inkling-2-Mini-MLX-2bit"
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default VisualInference/Inkling-2-Mini-MLX-2bit
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use VisualInference/Inkling-2-Mini-MLX-2bit with OpenClaw:
Start the MLX server
# Install MLX LM: uv tool install mlx-lm # Start a local OpenAI-compatible server: mlx_lm.server --model "VisualInference/Inkling-2-Mini-MLX-2bit"
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "VisualInference/Inkling-2-Mini-MLX-2bit" \ --custom-provider-id mlx-lm \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
File size: 16,722 Bytes
031489f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 | """Task-level checks through an OpenAI-compatible endpoint (the way an agent harness talks to the model).
Usage: task_eval.py <base_url> <model_id> <out.json> [--suites gsm8k,humaneval,json,tools] [--n-gsm 50] [--n-he 40]
Greedy decoding (temperature 0) so that variants are compared on the same footing.
"""
import json
import os
import re
import subprocess
import sys
import tempfile
import time
import urllib.request
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
from scenarios import TOOLS, run_tool # noqa: E402
ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
CONTROL = re.compile(r"<\|[a-z_0-9]+\|>")
def chat(base, model, messages, tools=None, max_tokens=800, effort="low", temperature=0.0, extra=None):
body = {"model": model, "messages": messages, "max_tokens": max_tokens, "temperature": temperature,
"reasoning_effort": effort}
if tools:
body["tools"] = tools
if extra:
body.update(extra)
req = urllib.request.Request(base.rstrip("/") + "/chat/completions", data=json.dumps(body).encode(),
headers={"content-type": "application/json"})
t0 = time.time()
with urllib.request.urlopen(req, timeout=900) as r:
data = json.load(r)
choice = data["choices"][0]
return choice["message"], choice.get("finish_reason"), data.get("usage", {}), time.time() - t0
def leaked(msg):
return bool(CONTROL.search((msg.get("content") or "")))
# ------------------------------------------------------------------ GSM8K
def number(text):
m = re.findall(r"Answer:\s*\$?\s*(-?[\d,]*\.?\d+)", text or "")
if not m:
m = re.findall(r"(-?[\d,]*\.?\d+)", text or "")
if not m:
return None
try:
return float(m[-1].replace(",", ""))
except ValueError:
return None
def suite_gsm8k(base, model, n):
import pyarrow.parquet as pq
rows = pq.read_table(os.path.join(ROOT, "data/evals/gsm8k_test.parquet")).to_pylist()
step = len(rows) // n
out = []
for i in range(n):
row = rows[i * step]
gold = float(row["answer"].split("####")[-1].strip().replace(",", ""))
msg, fin, usage, dt = chat(base, model, [{"role": "user", "content": row["question"] +
"\n\nSolve it, then give the final answer on the last line as 'Answer: <number>'."}],
max_tokens=900, effort="low")
pred = number(msg.get("content"))
out.append(dict(i=i * step, ok=pred is not None and abs(pred - gold) < 1e-4, pred=pred, gold=gold, finish=fin,
leaked=leaked(msg), tokens=usage.get("completion_tokens"), seconds=round(dt, 1)))
return out
# ------------------------------------------------------------------ HumanEval
SANDBOX = """(version 1)
(allow default)
(deny network*)
(deny file-write*)
(allow file-write* (subpath "{tmp}") (literal "/dev/null") (literal "/dev/dtracehelper"))
"""
def run_python(code, timeout=15):
with tempfile.TemporaryDirectory() as tmp:
tmp = os.path.realpath(tmp)
path = os.path.join(tmp, "prog.py")
open(path, "w").write(code)
try:
p = subprocess.run(["/usr/bin/sandbox-exec", "-p", SANDBOX.format(tmp=tmp), sys.executable, "-I", "-B", path],
capture_output=True, text=True, timeout=timeout, cwd=tmp)
return p.returncode == 0, (p.stderr or "")[-300:]
except subprocess.TimeoutExpired:
return False, "timeout"
def extract_code(text):
blocks = re.findall(r"```(?:python|py)?\n(.*?)```", text or "", flags=re.S)
return max(blocks, key=len) if blocks else (text or "")
def suite_humaneval(base, model, n):
import pyarrow.parquet as pq
rows = pq.read_table(os.path.join(ROOT, "data/evals/humaneval.parquet")).to_pylist()
step = len(rows) // n
out = []
for i in range(n):
row = rows[i * step]
msg, fin, usage, dt = chat(base, model, [{"role": "user", "content":
"Complete this Python function. Reply with the full function (including the signature "
"and any imports) in one ```python block and nothing else.\n\n" + row["prompt"]}],
max_tokens=900, effort="low")
code = extract_code(msg.get("content"))
ok, err = run_python(code + "\n\n" + row["test"] + f"\n\ncheck({row['entry_point']})\n")
out.append(dict(task=row["task_id"], ok=ok, err=None if ok else err, finish=fin, leaked=leaked(msg),
tokens=usage.get("completion_tokens"), seconds=round(dt, 1)))
return out
# ------------------------------------------------------------------ structured output
JSON_CASES = [
("Return a JSON object with keys \"city\" (string) and \"population_millions\" (number) for the capital of Japan. Output only JSON.",
lambda d: isinstance(d, dict) and d.get("city") == "Tokyo" and isinstance(d.get("population_millions"), (int, float))),
("Output only a JSON array of the first six prime numbers.", lambda d: d == [2, 3, 5, 7, 11, 13]),
("Extract as JSON {\"name\": string, \"email\": string, \"age\": integer} from: 'Contact Maria Okafor (34) at maria.okafor@example.org.' Output only JSON.",
lambda d: isinstance(d, dict) and d.get("name") == "Maria Okafor" and d.get("email") == "maria.okafor@example.org" and d.get("age") == 34),
("Output only JSON: an object mapping each of the words \"apple\", \"kiwi\", \"banana\" to its number of letters.",
lambda d: d == {"apple": 5, "kiwi": 4, "banana": 6}),
("Convert to JSON with keys \"items\" (array of {\"sku\": string, \"qty\": integer}) and \"rush\" (boolean): 'Two of SKU A-17 and five of SKU B-02, rush delivery please.' Output only JSON.",
lambda d: isinstance(d, dict) and d.get("rush") is True and sorted((x.get("sku"), x.get("qty")) for x in d.get("items", [])) == [("A-17", 2), ("B-02", 5)]),
("Output only JSON: {\"sorted\": [...]} with the numbers 42, 7, 19, 3 in ascending order.", lambda d: d == {"sorted": [3, 7, 19, 42]}),
("Return only a JSON object {\"sentiment\": \"positive\"|\"negative\"|\"neutral\", \"confidence\": number between 0 and 1} for: 'The update broke everything and support never answered.'",
lambda d: isinstance(d, dict) and d.get("sentiment") == "negative" and isinstance(d.get("confidence"), (int, float)) and 0 <= d["confidence"] <= 1),
("Output only JSON: {\"date\": \"YYYY-MM-DD\"} for the 4th of July 2031.", lambda d: d == {"date": "2031-07-04"}),
("Output only a JSON object with a key \"quote\" whose value is the exact text: She said \"stop\" and left.\\ Then nothing.",
lambda d: isinstance(d, dict) and d.get("quote") == 'She said "stop" and left.\\ Then nothing.'),
("Output only JSON: {\"tree\": {\"value\": 1, \"children\": [{\"value\": 2, \"children\": []}, {\"value\": 3, \"children\": [{\"value\": 4, \"children\": []}]}]}} but with every value doubled.",
lambda d: d == {"tree": {"value": 2, "children": [{"value": 4, "children": []}, {"value": 6, "children": [{"value": 8, "children": []}]}]}}),
]
def parse_json(text):
text = (text or "").strip()
m = re.search(r"```(?:json)?\n(.*?)```", text, flags=re.S)
if m:
text = m.group(1).strip()
return json.loads(text)
def suite_json(base, model, _n=None):
out = []
for i, (prompt, check) in enumerate(JSON_CASES):
msg, fin, usage, dt = chat(base, model, [{"role": "user", "content": prompt}], max_tokens=1200, effort="none")
try:
data = parse_json(msg.get("content")); valid = True
ok = bool(check(data))
except Exception: # noqa: BLE001
valid, ok = False, False
out.append(dict(i=i, ok=ok, valid_json=valid, finish=fin, leaked=leaked(msg), seconds=round(dt, 1),
content=None if ok else (msg.get("content") or "")[:200]))
return out
# ------------------------------------------------------------------ tool use
def args_of(call):
a = call["function"].get("arguments")
return json.loads(a) if isinstance(a, str) else a
SINGLE = [ # (tools, user, expected tool, argument check)
(["get_weather", "calculator"], "What's the weather in Tokyo in fahrenheit?", "get_weather",
lambda a: a.get("city", "").lower() == "tokyo" and a.get("unit") == "fahrenheit"),
(["get_weather", "web_search"], "Look up the weather for Cairo.", "get_weather", lambda a: a.get("city", "").lower() == "cairo"),
(["calculator", "web_search"], "Use the calculator to compute 48 * 365.", "calculator",
lambda a: re.sub(r"\s", "", a.get("expression", "")) in ("48*365", "365*48")),
(["terminal", "read_file", "write_file"], "List the files in the current directory.", "terminal", lambda a: "ls" in a.get("command", "")),
(["terminal", "read_file", "write_file"], "Show me the contents of app/config.py.", None,
lambda a: "app/config.py" in json.dumps(a)),
(["terminal", "read_file", "write_file"], "Create a file notes/todo.txt containing exactly the text: buy milk", None,
lambda a: (a.get("path", "").endswith("notes/todo.txt") and a.get("content", "").strip() == "buy milk")
or ("todo.txt" in a.get("command", "") and "buy milk" in a.get("command", ""))),
(["search_orders"], "Find customer C-88's shipped orders over 100 dollars that are tagged gift.", "search_orders",
lambda a: a.get("customer_id") == "C-88" and a.get("status") == "shipped" and (a.get("filters") or {}).get("min_total") == 100
and (a.get("filters") or {}).get("tags") == ["gift"]),
(["search_orders", "calculator"], "List open orders for customer C-19.", "search_orders",
lambda a: a.get("customer_id") == "C-19" and a.get("status") == "open"),
(["web_search", "get_weather"], "Search the web for the MLX array framework, three results at most.", "web_search",
lambda a: "mlx" in a.get("query", "").lower() and a.get("max_results", 3) <= 3),
(["terminal"], "Run the test suite with pytest in quiet mode.", "terminal", lambda a: "pytest" in a.get("command", "")),
(["write_file", "read_file"], "Write a file hello.py that prints the string He said \"hi\" on one line and a backslash \\ on the next.",
"write_file", lambda a: a.get("path", "").endswith("hello.py") and "print" in a.get("content", "")),
(["get_weather", "calculator", "web_search", "terminal", "read_file", "write_file", "search_orders"],
"Use the weather tool for Oslo in celsius.", "get_weather", lambda a: a.get("city", "").lower() == "oslo" and a.get("unit", "celsius") == "celsius"),
]
NO_TOOL = [
(["get_weather", "web_search"], "No lookup needed: what is 2 + 2?"),
(["terminal", "read_file"], "Just say hello in French, no tools."),
]
MULTI = [ # (tools, system, user, check(final_text, calls))
(["get_weather"], None, "Is it warmer in Lisbon or in Oslo right now? Use the tool.",
lambda text, calls: "lisbon" in text.lower() and len(calls) >= 2),
(["terminal", "read_file", "write_file"], "You are a coding agent in a sandboxed repository. Use the tools; keep answers short.",
"What is the value of RETRIES in app/config.py?", lambda text, calls: "3" in text and len(calls) >= 1),
(["terminal", "read_file", "write_file"], "You are a coding agent in a sandboxed repository. Use the tools; keep answers short.",
"Open docs/ARCHITECTURE.md and summarise it. If it does not exist, say so.",
lambda text, calls: any(w in text.lower() for w in ("not exist", "no such", "doesn't exist", "does not exist", "isn't", "missing", "no `docs", "not found", "there is no"))),
(["calculator"], None, "Use the calculator: what is (1234 * 57) - 999, and what is that divided by 7?",
lambda text, calls: "69339" in text.replace(",", "") and len(calls) >= 1),
(["search_orders"], None, "How many orders does customer C-88 have in total, and which is the most expensive?",
lambda text, calls: "A-1042" in text and len(calls) >= 1),
]
def suite_tools(base, model, _n=None):
import scenarios
out = []
for i, (names, user, want, check) in enumerate(SINGLE):
msg, fin, usage, dt = chat(base, model, [{"role": "user", "content": user}], tools=[TOOLS[t] for t in names],
max_tokens=500, effort="low")
calls = msg.get("tool_calls") or []
rec = dict(kind="single", i=i, n_calls=len(calls), finish=fin, leaked=leaked(msg), seconds=round(dt, 1))
try:
a = args_of(calls[0]) if calls else None
rec["valid_json"] = a is not None
rec["right_tool"] = bool(calls) and (want is None or calls[0]["function"]["name"] == want)
rec["ok"] = bool(rec["right_tool"] and check(a) and fin == "tool_calls")
except Exception as exc: # noqa: BLE001
rec.update(valid_json=False, right_tool=False, ok=False, error=str(exc)[:120])
if not rec["ok"]:
rec["got"] = json.dumps(calls)[:300] if calls else (msg.get("content") or "")[:200]
out.append(rec)
for i, (names, user) in enumerate(NO_TOOL):
msg, fin, usage, dt = chat(base, model, [{"role": "user", "content": user}], tools=[TOOLS[t] for t in names],
max_tokens=300, effort="low")
calls = msg.get("tool_calls") or []
out.append(dict(kind="no_tool", i=i, ok=not calls and bool((msg.get("content") or "").strip()) and fin == "stop",
n_calls=len(calls), finish=fin, leaked=leaked(msg), seconds=round(dt, 1)))
for i, (names, system, user, check) in enumerate(MULTI):
scenarios.FILES.update(ORIG_FILES)
for k in list(scenarios.FILES):
if k not in ORIG_FILES:
del scenarios.FILES[k]
messages = ([{"role": "system", "content": system}] if system else []) + [{"role": "user", "content": user}]
all_calls, rounds, leak, fin, t0, err = [], 0, False, None, time.time(), None
try:
for rounds in range(1, 13):
msg, fin, usage, dt = chat(base, model, messages, tools=[TOOLS[t] for t in names], max_tokens=900, effort="low")
leak = leak or leaked(msg)
calls = msg.get("tool_calls") or []
clean = {"role": "assistant", "content": msg.get("content") or ""}
if calls:
clean["tool_calls"] = calls
messages.append(clean)
if not calls:
break
all_calls += calls
for c in calls:
messages.append({"role": "tool", "tool_call_id": c.get("id"), "name": c["function"]["name"],
"content": run_tool(c["function"]["name"], args_of(c) or {})})
final = messages[-1].get("content") or ""
ok = bool(messages[-1]["role"] == "assistant" and check(final, all_calls))
except Exception as exc: # noqa: BLE001
ok, final, err = False, "", str(exc)[:200]
out.append(dict(kind="multi", i=i, ok=ok, rounds=rounds, n_calls=len(all_calls), finish=fin, leaked=leak,
seconds=round(time.time() - t0, 1), final=final[:240], error=err))
return out
import scenarios as _sc # noqa: E402
ORIG_FILES = dict(_sc.FILES)
SUITES = {"gsm8k": suite_gsm8k, "humaneval": suite_humaneval, "json": suite_json, "tools": suite_tools}
def main():
args = sys.argv[1:]
base, model, out_path = args[:3]
suites, n = ["gsm8k", "humaneval", "json", "tools"], {"gsm8k": 50, "humaneval": 40}
rest = args[3:]
while rest:
if rest[0] == "--suites":
suites = rest[1].split(","); rest = rest[2:]
elif rest[0] == "--n-gsm":
n["gsm8k"] = int(rest[1]); rest = rest[2:]
elif rest[0] == "--n-he":
n["humaneval"] = int(rest[1]); rest = rest[2:]
else:
raise SystemExit("unknown flag " + rest[0])
results = json.load(open(out_path)) if os.path.exists(out_path) else {}
for s in suites:
if s in results:
continue
t0 = time.time()
rows = SUITES[s](base, model, n.get(s))
results[s] = dict(passed=sum(r["ok"] for r in rows), total=len(rows), leaked=sum(bool(r.get("leaked")) for r in rows),
seconds=round(time.time() - t0), rows=rows)
json.dump(results, open(out_path, "w"), indent=1)
print(f"{s:10s} {results[s]['passed']}/{results[s]['total']} control-token leaks {results[s]['leaked']} {results[s]['seconds']}s", flush=True)
if __name__ == "__main__":
main()
|