Phi-4-mini-instruct-FP8

Phi4-mini model quantized with torchao float8 dynamic activation and float8 weight quantization (per row granularity), by PyTorch team. Use it directly, or serve using vLLM with 36% VRAM reduction (5.70 GB needed), 1.15x-1.2x speedup and little to no accuracy impact on H100.

Inference with vLLM

Install vllm nightly to get some recent changes:

pip install vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
pip install --pre torchao torch --index-url https://download.pytorch.org/whl/nightly/cu126

To use fbgemm kernels:

pip install --pre fbgemm-gpu-genai --index-url https://download.pytorch.org/whl/nightly/cu126

Code Example

from vllm import LLM, SamplingParams

# Sample prompts.
prompts = [
    "Hello, my name is",
    "The president of the United States is",
    "The capital of France is",
    "The future of AI is",
]
# Create a sampling params object.
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)


if __name__ == '__main__':
    # Create an LLM.
    llm = LLM(model="pytorch/Phi-4-mini-instruct-FP8")
    # Generate texts from the prompts.
    # The output is a list of RequestOutput objects
    # that contain the prompt, generated text, and other information.
    outputs = llm.generate(prompts, sampling_params)
    # Print the outputs.
    print("\nGenerated Outputs:\n" + "-" * 60)
    for output in outputs:
        prompt = output.prompt
        generated_text = output.outputs[0].text
        print(f"Prompt:    {prompt!r}")
        print(f"Output:    {generated_text!r}")
        print("-" * 60)

Note: please use VLLM_DISABLE_COMPILE_CACHE=1 to disable compile cache when running this code, e.g. VLLM_DISABLE_COMPILE_CACHE=1 python example.py, since there are some issues with the composability of compile in vLLM and torchao, this is expected be resolved in pytorch 2.8.

Serving

Then we can serve with the following command:

vllm serve pytorch/Phi-4-mini-instruct-FP8 --tokenizer microsoft/Phi-4-mini-instruct -O3

Inference with Transformers

Install the required packages:

# for compatibility with modeling file in checkpoint
pip install transformers==4.53.0
pip install --pre torchao torch --index-url https://download.pytorch.org/whl/nightly/cu126
pip install accelerate

To use fbgemm kernels:

pip install --pre fbgemm-gpu-genai --index-url https://download.pytorch.org/whl/nightly/cu126

Example:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
torch.random.manual_seed(0)

model_path = "pytorch/Phi-4-mini-instruct-FP8"

model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_path)
 
messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"},
    {"role": "assistant", "content": "Sure! Here are some ways to eat bananas and dragonfruits together: 1. Banana and dragonfruit smoothie: Blend bananas and dragonfruits together with some milk and honey. 2. Banana and dragonfruit salad: Mix sliced bananas and dragonfruits together with some lemon juice and honey."},
    {"role": "user", "content": "What about solving an 2x + 3 = 7 equation?"},
]
 
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
)
 
generation_args = {
    "max_new_tokens": 500,
    "return_full_text": False,
    "temperature": 0.0,
    "do_sample": False,
}
 
output = pipe(messages, **generation_args)
print(output[0]['generated_text'])

Quantization Recipe

Install the required packages:

# for compatibility with modeling file in checkpoint
pip install transformers==4.53.0
pip install --pre torchao --index-url https://download.pytorch.org/whl/nightly/cu126
pip install torch
pip install accelerate

Use the following code to get the quantized model:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

model_id = "microsoft/Phi-4-mini-instruct"

from torchao.quantization import Float8DynamicActivationFloat8WeightConfig, PerRow
quant_config = Float8DynamicActivationFloat8WeightConfig(granularity=PerRow())
quantization_config = TorchAoConfig(quant_type=quant_config)
quantized_model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.bfloat16, quantization_config=quantization_config)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Push to hub
USER_ID = "YOUR_USER_ID"
MODEL_NAME = model_id.split("/")[-1]
save_to = f"{USER_ID}/{MODEL_NAME}-FP8"
quantized_model.push_to_hub(save_to, safe_serialization=False)
tokenizer.push_to_hub(save_to)

# Manual Testing
prompt = "Hey, are you conscious? Can you talk to me?"
messages = [
    {
        "role": "system",
        "content": "",
    },
    {"role": "user", "content": prompt},
]
templated_prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
print("Prompt:", prompt)
print("Templated prompt:", templated_prompt)
inputs = tokenizer(
    templated_prompt,
    return_tensors="pt",
).to("cuda")
generated_ids = quantized_model.generate(**inputs, max_new_tokens=128)
output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print("Response:", output_text[0][len(prompt):])

Note: to push_to_hub you need to run

pip install -U "huggingface_hub[cli]"
huggingface-cli login

and use a token with write access, from https://huggingface.co/settings/tokens

Model Quality

We rely on lm-evaluation-harness to evaluate the quality of the quantized model.

<details> <summary> Reproduce Model Quality Results </summary>

Need to install lm-eval from source: https://github.com/EleutherAI/lm-evaluation-harness#install

baseline

lm_eval --model hf --model_args pretrained=microsoft/Phi-4-mini-instruct --tasks hellaswag --device cuda:0 --batch_size 8

float8 dynamic activation and float8 weight quantization (FP8)

lm_eval --model hf --model_args pretrained=pytorch/Phi-4-mini-instruct-FP8 --tasks hellaswag --device cuda:0 --batch_size 8

</details>

Peak Memory Usage

Results

<details> <summary> Reproduce Peak Memory Usage Results </summary>

We can use the following code to get a sense of peak memory usage during inference:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

# use "microsoft/Phi-4-mini-instruct" or "pytorch/Phi-4-mini-instruct-FP8"
model_id = "pytorch/Phi-4-mini-instruct-FP8"
quantized_model = AutoModelForCausalLM.from_pretrained(model_id, device_map="cuda:0", torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained(model_id)

torch.cuda.reset_peak_memory_stats()

prompt = "Hey, are you conscious? Can you talk to me?"
messages = [
    {
        "role": "system",
        "content": "",
    },
    {"role": "user", "content": prompt},
]
templated_prompt = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
)
print("Prompt:", prompt)
print("Templated prompt:", templated_prompt)
inputs = tokenizer(
    templated_prompt,
    return_tensors="pt",
).to("cuda")
generated_ids = quantized_model.generate(**inputs, max_new_tokens=128)
output_text = tokenizer.batch_decode(
    generated_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
)
print("Response:", output_text[0][len(prompt):])

mem = torch.cuda.max_memory_reserved() / 1e9
print(f"Peak Memory Usage: {mem:.02f} GB")

</details>

Model Performance

Results (H100 machine)

Note the result of latency (benchmark_latency) is in seconds, and serving (benchmark_serving) is in number of requests per second.

<details> <summary> Reproduce Model Performance Results </summary>

Setup

Get vllm source code:

git clone [email protected]:vllm-project/vllm.git

Install vllm

VLLM_USE_PRECOMPILED=1 pip install --editable .

To use fbgemm kernels:

pip install fbgemm-gpu-genai

Run the benchmarks under vllm root folder:

benchmark_latency

baseline

vllm bench latency --input-len 256 --output-len 256 --model microsoft/Phi-4-mini-instruct --batch-size 1

FP8

VLLM_DISABLE_COMPILE_CACHE=1 vllm bench latency --input-len 256 --output-len 256 --model pytorch/Phi-4-mini-instruct-FP8 --batch-size 1

</details>

Paper: TorchAO: PyTorch-Native Training-to-Serving Model Optimization

The model's quantization is powered by TorchAO, a framework presented in the paper TorchAO: PyTorch-Native Training-to-Serving Model Optimization.

Abstract: We present TorchAO, a PyTorch-native model optimization framework leveraging quantization and sparsity to provide an end-to-end, training-to-serving workflow for AI models. TorchAO supports a variety of popular model optimization techniques, including FP8 quantized training, quantization-aware training (QAT), post-training quantization (PTQ), and 2:4 sparsity, and leverages a novel tensor subclass abstraction to represent a variety of widely-used, backend agnostic low precision data types, including INT4, INT8, FP8, MXFP4, MXFP6, and MXFP8. TorchAO integrates closely with the broader ecosystem at each step of the model optimization pipeline, from pre-training (TorchTitan) to fine-tuning (TorchTune, Axolotl) to serving (HuggingFace, vLLM, SGLang, ExecuTorch), connecting an otherwise fragmented space in a single, unified workflow. TorchAO has enabled recent launches of the quantized Llama 3.2 1B/3B and LlamaGuard3-8B models and is open-source at this https URL .

Resources

• Official TorchAO GitHub Repository: https://github.com/pytorch/ao
• TorchAO Documentation: https://docs.pytorch.org/ao/stable/index.html

Disclaimer

PyTorch has not performed safety evaluations or red teamed the quantized models. Performance characteristics, outputs, and behaviors may differ from the original models. Users are solely responsible for selecting appropriate use cases, evaluating and mitigating for accuracy, safety, and fairness, ensuring security, and complying with all applicable laws and regulations.

Nothing contained in this Model Card should be interpreted as or deemed a restriction or modification to the licenses the models are released under, including any limitations of liability or disclaimers of warranties provided therein.