Text Generation
Transformers
Safetensors
lfm2
liquid
lfm2.5
edge
parallel-constrained-decoding
structured-generation
classification
inference-only
modal
conversational
Instructions to use monotykamary/LFM2.5-2.6B-RLCD with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use monotykamary/LFM2.5-2.6B-RLCD with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="monotykamary/LFM2.5-2.6B-RLCD") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("monotykamary/LFM2.5-2.6B-RLCD") model = AutoModelForCausalLM.from_pretrained("monotykamary/LFM2.5-2.6B-RLCD", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use monotykamary/LFM2.5-2.6B-RLCD with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "monotykamary/LFM2.5-2.6B-RLCD" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "monotykamary/LFM2.5-2.6B-RLCD", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/monotykamary/LFM2.5-2.6B-RLCD
- SGLang
How to use monotykamary/LFM2.5-2.6B-RLCD with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "monotykamary/LFM2.5-2.6B-RLCD" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "monotykamary/LFM2.5-2.6B-RLCD", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "monotykamary/LFM2.5-2.6B-RLCD" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "monotykamary/LFM2.5-2.6B-RLCD", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use monotykamary/LFM2.5-2.6B-RLCD with Docker Model Runner:
docker model run hf.co/monotykamary/LFM2.5-2.6B-RLCD
Download scripts/build_pcd_release.py from monotykamary/LFM2.5-2.6B-RLCD: direct link, hf CLI and curl.
- Browser
- Download file 2.67 kB
-
https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD/resolve/main/scripts/build_pcd_release.py
- Command line
-
hf download hf://monotykamary/LFM2.5-2.6B-RLCD/scripts/build_pcd_release.py
-
curl -L -o build_pcd_release.py https://huggingface.co/monotykamary/LFM2.5-2.6B-RLCD/resolve/main/scripts/build_pcd_release.py
2.67 kB
| """Stage an explicit allowlist of code/docs/results; never copy credentials or .git.""" | |
| import argparse | |
| import hashlib | |
| import json | |
| import shutil | |
| from pathlib import Path | |
| ROOT = Path(__file__).resolve().parents[1] | |
| def build(reports): | |
| output = ROOT / "release" / "pcd" | |
| output.mkdir(parents=True, exist_ok=True) | |
| files = [ | |
| ROOT / name | |
| for name in ( | |
| "requirements-pcd.txt", | |
| "lfm25_pcd_modal.py", | |
| "README-PCD.md", | |
| "LICENSE-CODE", | |
| "THIRD_PARTY_NOTICES.md", | |
| "ruff.toml", | |
| ) | |
| ] | |
| files += sorted((ROOT / "pcd").glob("*.py")) | |
| files += sorted((ROOT / "tests" / "pcd").glob("*.py")) | |
| files += sorted((ROOT / "scripts").glob("*pcd*.py")) | |
| files += [ROOT / "docs" / "pcd-results.md"] | |
| files += [ROOT / name for name in reports] | |
| hashes = {} | |
| for source in files: | |
| relative = source.relative_to(ROOT) | |
| if not source.is_file() or source.is_symlink(): | |
| raise ValueError(f"expected a regular release file: {relative}") | |
| target = output / relative | |
| target.parent.mkdir(parents=True, exist_ok=True) | |
| shutil.copyfile(source, target) | |
| hashes[str(relative)] = hashlib.sha256(target.read_bytes()).hexdigest() | |
| manifest = { | |
| "files": hashes, | |
| "reports": reports, | |
| "core_files": [ | |
| "engine.py", | |
| "cache.py", | |
| "schema.py", | |
| "prompting.py", | |
| "config.py", | |
| "tasks.py", | |
| ], | |
| } | |
| for name in reports: | |
| report = json.loads((ROOT / name).read_text()) | |
| if not report.get("validation", {}).get("passed") or report.get("budget_exhausted"): | |
| raise ValueError(f"incomplete/failed release report: {name}") | |
| if report["task"] == "benchmark": | |
| for core in manifest["core_files"]: | |
| if report["source_sha256"][core] != hashes[f"pcd/{core}"]: | |
| raise ValueError(f"benchmark/code mismatch: {core}") | |
| # Fail on stale staging files rather than accidentally uploading leftovers. | |
| actual = {str(p.relative_to(output)) for p in output.rglob("*") if p.is_file()} | |
| unexpected = actual - set(hashes) - {"ARTIFACTS.json"} | |
| if unexpected: | |
| raise ValueError(f"unexpected files in release staging: {sorted(unexpected)}") | |
| (output / "ARTIFACTS.json").write_text(json.dumps(manifest, indent=2) + "\n") | |
| print(json.dumps({"directory": str(output), "files": len(hashes)}, indent=2)) | |
| if __name__ == "__main__": | |
| parser = argparse.ArgumentParser() | |
| parser.add_argument("reports", nargs="+") | |
| build(parser.parse_args().reports) | |