Safetensors
GGUF
English
Russian
lfm3
reasoning
test-time-compute
inference-time-scaling
code-generation
llama.cpp
negative-results
conversational
Instructions to use HelioAI/LFM3-1.7B-inference-modules-experiments with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M # Run inference directly in the terminal: llama cli -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M # Run inference directly in the terminal: llama cli -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M # Run inference directly in the terminal: ./build/bin/llama cli -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Use Docker
docker model run hf.co/HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
- Llama new
- LM Studio
- Jan
- Ollama
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with Ollama:
ollama run hf.co/HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
- Unsloth Desktop
- Pi
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with Docker Model Runner:
docker model run hf.co/HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
- Lemonade
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Run and chat with the model
lemonade run user.LFM3-1.7B-inference-modules-experiments-Q5_K_M
List all available models
lemonade list
- Hermes Agent
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
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 HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use HelioAI/LFM3-1.7B-inference-modules-experiments with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M
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 "HelioAI/LFM3-1.7B-inference-modules-experiments:Q5_K_M" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
File size: 7,843 Bytes
67ebcef | 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 | """Turns an answer written with a toolkit (kits/) into one self-contained file.
three : `import {...} from 'kit3d'` -> reads window.K3; the kit3d module is inserted before the page's module
script and the import map gets 'three' and 'three/addons/' if they are missing;
canvas : the kit2d script (global Kit) is inserted before the page's first script;
html : <link href="ui.css"> / <script src="ui.js"> are replaced by the inline kit (added if missing);
svg : the answer is a Python program using svgkit; it is run and the SVG it prints becomes the answer.
finalize(answer_text, cat) -> (text with one ```svg / ```html block, problem or None, source used for rendering)
"""
import json
import os
import re
import shutil
import subprocess
import sys
import tempfile
import xml.dom.minidom
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
import sandbox # noqa: E402
from grade import extract_code # noqa: E402
from render import extract # noqa: E402
from vis_autofix import fix_importmap # noqa: E402
KITS = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "kits")
def _read(name):
return open(os.path.join(KITS, name), encoding="utf-8").read()
KIT3D, KIT2D, UICSS, UIJS = (_read(n) for n in ("kit3d.js", "kit2d.js", "ui.css", "ui.js"))
IMPORT_KIT3D = re.compile(r"""import\s+([^;'"]+?)\s+from\s+['"](?:\./)?kit3d(?:\.js)?['"]\s*;?""")
THREE_URL = "https://unpkg.com/three@0.160.0/build/three.module.js"
ADDONS_URL = "https://unpkg.com/three@0.160.0/examples/jsm/"
def _destructure(clause):
"""'{ a, b as c }' | '* as K' | 'K' | 'K, { a }' -> statements that read window.K3."""
clause, out = clause.strip(), []
m = re.fullmatch(r"\*\s+as\s+(\w+)", clause)
if m:
return f"const {m.group(1)} = window.K3;"
d = re.match(r"(\w+)\s*(,|$)", clause)
if d:
out.append(f"const {d.group(1)} = window.K3;")
b = re.search(r"\{(.*?)\}", clause, re.S)
if b:
names = []
for part in b.group(1).split(","):
part = part.strip()
if part:
a = re.fullmatch(r"(\w+)\s+as\s+(\w+)", part)
names.append(f"{a.group(1)}: {a.group(2)}" if a else part)
out.append("const { " + ", ".join(names) + " } = window.K3;")
return " ".join(out)
def inline_three(src):
if not IMPORT_KIT3D.search(src):
return fix_importmap(src)
src = IMPORT_KIT3D.sub(lambda m: _destructure(m.group(1)), src)
src = fix_importmap(src)
m = re.search(r"""<script\s+type=["']importmap["']\s*>(.*?)</script>""", src, re.S)
if m:
try:
j = json.loads(m.group(1))
imp = j.setdefault("imports", {})
if "three" not in imp or "three/addons/" not in imp:
imp.setdefault("three", THREE_URL)
imp.setdefault("three/addons/", ADDONS_URL)
src = src[:m.start(1)] + json.dumps(j) + src[m.end(1):]
except ValueError:
pass
kit = '<script type="module">\n' + KIT3D + "\n</script>\n"
m = re.search(r"""<script\s+type=["']module["']""", src)
if not m:
m = re.search(r"</body>", src, re.I)
return src[:m.start()] + kit + src[m.start():] if m else src + kit
def inline_canvas(src):
kit = "<script>\n" + KIT2D + "\n</script>\n"
tag = re.search(r"""<script[^>]*\bsrc=["'](?:\./)?kit2d(?:\.js)?["'][^>]*>\s*</script>""", src, re.I)
if tag:
return src[:tag.start()] + kit + src[tag.end():]
if "Kit." not in src or "window.Kit" in src:
return src
m = re.search(r"<script", src, re.I) or re.search(r"</body>", src, re.I)
return src[:m.start()] + kit + src[m.start():] if m else src + kit
def inline_ui(src):
css, js = "<style>\n" + UICSS + "\n</style>", "<script>\n" + UIJS + "\n</script>"
link = re.search(r"""<link[^>]*href=["'](?:\./)?ui\.css["'][^>]*>""", src, re.I)
if link:
src = src[:link.start()] + css + src[link.end():]
else: # the kit comes first so the page's own <style> can override it
m = re.search(r"<head[^>]*>", src, re.I)
src = src[:m.end()] + "\n" + css + src[m.end():] if m else css + src
tags = list(re.finditer(r"""<script[^>]*\bsrc=["'](?:\./)?ui\.js["'][^>]*>\s*</script>""", src, re.I))
if tags: # the first reference becomes the inline kit, repeated references are dropped
for t in reversed(tags[1:]):
src = src[:t.start()] + src[t.end():]
return src[:tags[0].start()] + js + src[tags[0].end():]
m = re.search(r"</body>", src, re.I)
return src[:m.start()] + js + "\n" + src[m.start():] if m else src + js
def run_svg_program(code, timeout=20):
"""Runs a svgkit program in a fresh folder; returns (svg or None, problem or None)."""
d = tempfile.mkdtemp(prefix="svgkit_")
try:
shutil.copy(os.path.join(KITS, "svgkit.py"), d)
path = os.path.join(d, "draw.py")
with open(path, "w", encoding="utf-8") as f:
f.write(code)
try:
r = sandbox.run([sys.executable, "-E", "-s", path], input=b"", capture_output=True, timeout=timeout, cwd=d,
creationflags=getattr(subprocess, "CREATE_NO_WINDOW", 0))
except subprocess.TimeoutExpired:
return None, f"The program did not finish within {timeout} seconds."
out, err = r.stdout.decode(errors="replace"), r.stderr.decode(errors="replace")
if r.returncode != 0:
return None, "The program failed:\n" + "\n".join(err.strip().splitlines()[-6:])
finally:
shutil.rmtree(d, ignore_errors=True)
m = re.search(r"<svg.*</svg>", out, re.S)
if not m:
return None, "The program printed no <svg>...</svg> document (end it with print(s)). It printed:\n" + out[:300]
try:
xml.dom.minidom.parseString(m.group(0).encode("utf-8"))
except Exception as e:
return None, f"The printed SVG is not valid XML: {str(e)[:200]}"
return m.group(0), None
def finalize(text, cat):
"""(final answer text, problem or None, source for rendering or None)."""
if cat == "svg":
code = extract_code(text)
if code is None or "svgkit" not in code:
src = extract(text, "svg") # a direct SVG answer is still accepted
if src:
return f"```svg\n{src}\n```", None, src
return text, "Your answer does not contain the Python program in a ```python code block.", None
svg, problem = run_svg_program(code)
return (f"```svg\n{svg}\n```" if svg else text), problem, svg
src = extract(text, "html")
if not src:
return text, None, None
new = {"three": inline_three, "canvas": inline_canvas, "html": inline_ui}[cat](src)
return text.replace(src, new), None, new
CPLIB = _read("cplib.py").split('"""', 2)[2].lstrip() # the library without its module docstring
CPLIB_IMPORT = re.compile(r"^[ \t]*(from\s+cplib\s+import\s+[^\n]+|import\s+cplib\b[^\n]*)$", re.M)
def inline_cplib(text):
"""Replaces `from cplib import ...` / `import cplib` in the answer's last ```python block by the library."""
blocks = list(re.finditer(r"```(?:python|py|Python)?[ \t]*\n(.*?)```", text or "", re.S))
if not blocks or not CPLIB_IMPORT.search(blocks[-1].group(1)):
return text
code = blocks[-1].group(1)
module_style = re.search(r"^[ \t]*import\s+cplib\b", code, re.M)
lib = CPLIB + ("\nimport types as _types\ncplib = _types.SimpleNamespace(**{k: v for k, v in dict(globals()).items() "
"if not k.startswith('_')})\n" if module_style else "")
new = CPLIB_IMPORT.sub(lambda m: "# --- cplib (inlined) ---\n" + lib + "# --- end of cplib ---", code, count=1)
new = CPLIB_IMPORT.sub("", new)
s, e = blocks[-1].span(1)
return text[:s] + new + text[e:]
|