MiniCPM5-2B-FP8

MiniCPM5-2B — FP8 (compressed-tensors)

FP8-dynamic quantization of openbmb/MiniCPM5-2B, weights and activations in float8_e4m3 (per-channel weights, per-token dynamic activations), lm_head and embeddings kept in bf16. Produced with llm-compressor 0.13.

• 2.84 GiB on disk (bf16 base is 4.68 GiB — 39 % smaller)
• Serves on vLLM (compressed-tensors), native FP8 tensor-core path on Ada / Blackwell
• Near-lossless on coding benchmarks — see the eval below

This is the recommended quant for MiniCPM5-2B when you want maximum quality retention. If you need to fit closer to 2 GB, see the NVFP4 W4A16 build (smaller, ~2.4 pp HumanEval / ~4.8 pp MBPP cost).

Evaluation

lm-evaluation-harness, vLLM 0.26 backend, greedy decoding, median of 3 draws with the range (the harness is non-deterministic run-to-run even at greedy — a single draw is not a reproducible score). All arms measured in one session against the released checkpoint. HumanEval-instruct pass@1/create_test (n = 164); MBPP base 3-shot (n = 500).

Both FP8 deltas sit inside the eval's own run-to-run spread — FP8 is effectively lossless here. NVFP4 costs real accuracy on a 2.5 B dense model (little redundancy to absorb 4-bit weights); pick it only if the extra ~0.8 GiB matters for your KV budget.

Full comparison, serving notes, and a fine-tune experiment that regressed the base's coding (so these quants target the released checkpoint): https://github.com/t-timms/minicpm5-2b-quants

Usage

vllm serve Ttimms/MiniCPM5-2B-FP8 --max-model-len 32768 --kv-cache-dtype fp8

from vllm import LLM, SamplingParams
llm = LLM("Ttimms/MiniCPM5-2B-FP8", trust_remote_code=True)
print(llm.chat([{"role": "user", "content": "Write a Python LRU cache."}],
               SamplingParams(temperature=0.6, max_tokens=512))[0].outputs[0].text)

On WSL, set VLLM_USE_V2_MODEL_RUNNER=0 (the V2 runner needs UVA, which WSL disables).

Method & provenance

• Quantizer: llm-compressor 0.13, QuantizationModifier(scheme="FP8_DYNAMIC"), ignore lm_head + embed_tokens. No calibration data required (dynamic activations).
• Base: openbmb/MiniCPM5-2B (LlamaForCausalLM, 2.5 B, Apache-2.0).
• Built and evaluated on an RTX 5070 Ti (Blackwell, SM120), vLLM 0.26.

License

Apache-2.0, inherited from openbmb/MiniCPM5-2B.