Text Generation
Transformers
Safetensors
English
Chinese
llama
minicpm
minicpm5
fp8
compressed-tensors
quantization
vllm
blackwell
code
on-device
conversational
text-generation-inference
Instructions to use Ttimms/MiniCPM5-2B-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Ttimms/MiniCPM5-2B-FP8 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Ttimms/MiniCPM5-2B-FP8") 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("Ttimms/MiniCPM5-2B-FP8") model = AutoModelForCausalLM.from_pretrained("Ttimms/MiniCPM5-2B-FP8", 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 Ttimms/MiniCPM5-2B-FP8 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Ttimms/MiniCPM5-2B-FP8" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Ttimms/MiniCPM5-2B-FP8", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Ttimms/MiniCPM5-2B-FP8
- SGLang
How to use Ttimms/MiniCPM5-2B-FP8 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 "Ttimms/MiniCPM5-2B-FP8" \ --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": "Ttimms/MiniCPM5-2B-FP8", "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 "Ttimms/MiniCPM5-2B-FP8" \ --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": "Ttimms/MiniCPM5-2B-FP8", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Ttimms/MiniCPM5-2B-FP8 with Docker Model Runner:
docker model run hf.co/Ttimms/MiniCPM5-2B-FP8
docs: full quant comparison table + serving notes + writeup link
Browse files
README.md
CHANGED
|
@@ -36,25 +36,32 @@ activations), `lm_head` and embeddings kept in bf16. Produced with
|
|
| 36 |
This is the recommended quant for MiniCPM5-2B when you want maximum quality
|
| 37 |
retention. If you need to fit closer to 2 GB, see the
|
| 38 |
[NVFP4 W4A16 build](https://huggingface.co/Ttimms/MiniCPM5-2B-NVFP4)
|
| 39 |
-
(smaller, ~2
|
| 40 |
|
| 41 |
## Evaluation
|
| 42 |
|
| 43 |
-
`lm-evaluation-harness`, vLLM backend, greedy decoding, **3 draws**
|
| 44 |
-
harness is non-deterministic run-to-run even at greedy —
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
|
| 50 |
-
|
|
| 51 |
-
|
|
| 52 |
-
|
| 53 |
-
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
## Usage
|
| 60 |
|
|
@@ -69,14 +76,15 @@ print(llm.chat([{"role": "user", "content": "Write a Python LRU cache."}],
|
|
| 69 |
SamplingParams(temperature=0.6, max_tokens=512))[0].outputs[0].text)
|
| 70 |
```
|
| 71 |
|
|
|
|
|
|
|
|
|
|
| 72 |
## Method & provenance
|
| 73 |
|
| 74 |
- Quantizer: `llm-compressor` 0.13, `QuantizationModifier(scheme="FP8_DYNAMIC")`,
|
| 75 |
ignore `lm_head` + `embed_tokens`. No calibration data required (dynamic activations).
|
| 76 |
- Base: `openbmb/MiniCPM5-2B` (`LlamaForCausalLM`, 2.5 B, Apache-2.0).
|
| 77 |
- Built and evaluated on an RTX 5070 Ti (Blackwell, SM120), vLLM 0.26.
|
| 78 |
-
- Full quant-format comparison + methodology:
|
| 79 |
-
<https://github.com/t-timms/blackwell-16gb-moe>
|
| 80 |
|
| 81 |
## License
|
| 82 |
|
|
|
|
| 36 |
This is the recommended quant for MiniCPM5-2B when you want maximum quality
|
| 37 |
retention. If you need to fit closer to 2 GB, see the
|
| 38 |
[NVFP4 W4A16 build](https://huggingface.co/Ttimms/MiniCPM5-2B-NVFP4)
|
| 39 |
+
(smaller, ~2.4 pp HumanEval / ~4.8 pp MBPP cost).
|
| 40 |
|
| 41 |
## Evaluation
|
| 42 |
|
| 43 |
+
`lm-evaluation-harness`, vLLM 0.26 backend, greedy decoding, **median of 3 draws**
|
| 44 |
+
with the range (the harness is non-deterministic run-to-run even at greedy — a
|
| 45 |
+
single draw is not a reproducible score). All arms measured in one session against
|
| 46 |
+
the released checkpoint. HumanEval-instruct `pass@1`/`create_test` (n = 164);
|
| 47 |
+
MBPP base 3-shot (n = 500).
|
| 48 |
+
|
| 49 |
+
| build | HumanEval-inst | MBPP (3-shot) | size | Δ HE / MBPP |
|
| 50 |
+
|---|---:|---:|---:|---:|
|
| 51 |
+
| bf16 base | 86.59 % (85.98–86.59) | 50.60 % (50.40–51.00) | 4.68 GiB | — |
|
| 52 |
+
| **FP8-dynamic (this)** | **84.76 %** (84.15–85.37) | **48.80 %** (48.80–49.00) | **2.84 GiB** | **−1.8 / −1.8 pp** |
|
| 53 |
+
| NVFP4-W4A16 GPTQ | 84.15 % (81.71–84.15) | 45.80 % (45.60–46.40) | 2.03 GiB | −2.4 / −4.8 pp |
|
| 54 |
+
| NVFP4-W4A16 RTN | 79.88 % (77.44–79.88) | 41.20 % (41.20–41.80) | 2.03 GiB | −6.7 / −9.4 pp |
|
| 55 |
+
| mixed (MLP-NVFP4 + attn-FP8) | 84.15 % (82.93–85.98) | 46.60 % (46.00–46.80) | 2.20 GiB | −2.4 / −4.0 pp |
|
| 56 |
+
|
| 57 |
+
Both FP8 deltas sit inside the eval's own run-to-run spread — FP8 is effectively
|
| 58 |
+
lossless here. NVFP4 costs real accuracy on a 2.5 B dense model (little redundancy
|
| 59 |
+
to absorb 4-bit weights); pick it only if the extra ~0.8 GiB matters for your KV
|
| 60 |
+
budget.
|
| 61 |
+
|
| 62 |
+
Full comparison, serving notes, and a fine-tune experiment that regressed the
|
| 63 |
+
base's coding (so these quants target the released checkpoint):
|
| 64 |
+
**<https://github.com/t-timms/minicpm5-2b-quants>**
|
| 65 |
|
| 66 |
## Usage
|
| 67 |
|
|
|
|
| 76 |
SamplingParams(temperature=0.6, max_tokens=512))[0].outputs[0].text)
|
| 77 |
```
|
| 78 |
|
| 79 |
+
On WSL, set `VLLM_USE_V2_MODEL_RUNNER=0` (the V2 runner needs UVA, which WSL
|
| 80 |
+
disables).
|
| 81 |
+
|
| 82 |
## Method & provenance
|
| 83 |
|
| 84 |
- Quantizer: `llm-compressor` 0.13, `QuantizationModifier(scheme="FP8_DYNAMIC")`,
|
| 85 |
ignore `lm_head` + `embed_tokens`. No calibration data required (dynamic activations).
|
| 86 |
- Base: `openbmb/MiniCPM5-2B` (`LlamaForCausalLM`, 2.5 B, Apache-2.0).
|
| 87 |
- Built and evaluated on an RTX 5070 Ti (Blackwell, SM120), vLLM 0.26.
|
|
|
|
|
|
|
| 88 |
|
| 89 |
## License
|
| 90 |
|