---
library_name: transformers
tags:
- torchao
- phi
- phi4
- nlp
- code
- math
- chat
- conversational
license: mit
language:
- multilingual
base_model:
- microsoft/Phi-4-mini-instruct
pipeline_tag: text-generation
---

[Phi4-mini](https://huggingface.co/microsoft/Phi-4-mini-instruct) model quantized with [torchao](https://huggingface.co/docs/transformers/main/en/quantization/torchao) int4 weight only quantization with gemlite kernels, by PyTorch team.

# Installation
```
pip install transformers
pip install --pre torchao --index-url https://download.pytorch.org/whl/nightly/cu126
pip install git@github.com:EleutherAI/lm-evaluation-harness.git
pip install vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
pip install git+https://github.com/mobiusml/gemlite/
```

# Quantization Recipe


We used 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 GemliteUIntXWeightOnlyConfig
quant_config = GemliteUIntXWeightOnlyConfig()
quantization_config = TorchAoConfig(quant_type=quant_config)
quantized_model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", torch_dtype=torch.float16, quantization_config=quantization_config)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Push to hub
USER_ID = "YOUR_USER_ID"
save_to = f"{USER_ID}/{model_id}-int4wo-gemlite"
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?"
inputs = tokenizer(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(output_text)

# Local Benchmark
import torch.utils.benchmark as benchmark
from torchao.utils import benchmark_model
import torchao

def benchmark_fn(f, *args, **kwargs):
    # Manual warmup
    for _ in range(2):
        f(*args, **kwargs)

    t0 = benchmark.Timer(
        stmt="f(*args, **kwargs)",
        globals={"args": args, "kwargs": kwargs, "f": f},
        num_threads=torch.get_num_threads(),
    )
    return f"{(t0.blocked_autorange().mean):.3f}"

torchao.quantization.utils.recommended_inductor_config_setter()
quantized_model = torch.compile(quantized_model, mode="max-autotune")
print(f"{save_to} model:", benchmark_fn(quantized_model.generate, **inputs, max_new_tokens=128))
```
# Model Quality
We rely on [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness) to evaluate the quality of the quantized model.

## Installing the nightly version to get most recent updates
```
pip install git+https://github.com/EleutherAI/lm-evaluation-harness
```

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

## int4wo-gemlite
```
lm_eval --model hf --model_args pretrained=jerryzh168/phi4-mini-int4wo-gemlite --tasks hellaswag --device cuda:0 --batch_size 8
```

`TODO: more complete eval results`


| Benchmark                        |             |                   |
|----------------------------------|-------------|-------------------|
|                                  | Phi-4 mini-Ins | phi4-mini-int4wo-gemlite | 
| **Popular aggregated benchmark** |             |                   |
| **Reasoning**                    |             |                   |
| HellaSwag                        | 54.57        | 53.51            |
| **Multilingual**                 |             |                   |
| **Math**                         |             |                   |
| **Overall**                      | **TODO**    | **TODO**          |
 
# Model Performance

Our int4wo is only optimized for batch size 1, so we'll only benchmark the batch size 1 performance with vllm.
For batch size N, please see our [gemlite checkpoint](https://huggingface.co/jerryzh168/phi4-mini-int4wo-gemlite).

## Download vllm source code and install vllm
```
git clone git@github.com:vllm-project/vllm.git
VLLM_USE_PRECOMPILED=1 pip install .
```

## Download dataset
Download sharegpt dataset: `wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json`

Other datasets can be found in: https://github.com/vllm-project/vllm/tree/main/benchmarks
## benchmark_latency

Run the following under `vllm` source code root folder:

### baseline
```
python benchmarks/benchmark_latency.py --input-len 256 --output-len 256 --model microsoft/Phi-4-mini-instruct --batch-size 1
```

### int4wo-gemlite
```
python benchmarks/benchmark_latency.py --input-len 256 --output-len 256 --model jerryzh168/phi4-mini-int4wo-gemlite --batch-size 1
```

## benchmark_serving

We also benchmarked the throughput in a serving environment.

Run the following under `vllm` source code root folder:

### baseline
Server:
```
vllm serve microsoft/Phi-4-mini-instruct --tokenizer microsoft/Phi-4-mini-instruct -O3
```

Client:
```
python benchmarks/benchmark_serving.py --backend vllm --dataset-name sharegpt --tokenizer microsoft/Phi-4-mini-instruct --dataset-path ./ShareGPT_V3_unfiltered_cleaned_split.json --model microsoft/Phi-4-mini-instruct --num-prompts 1
```

### int4wo-gemlite
Server:
```
vllm serve jerryzh168/phi4-mini-int4wo-gemlite --tokenizer microsoft/Phi-4-mini-instruct -O3
```

Client:
```
python benchmarks/benchmark_serving.py --backend vllm --dataset-name sharegpt --tokenizer microsoft/Phi-4-mini-instruct --dataset-path ./ShareGPT_V3_unfiltered_cleaned_split.json --model jerryzh168/phi4-mini-int4wo-hqq --num-prompts 1
```

# Serving with vllm
We can use the same command we used in serving benchmarks to serve the model with vllm
```
vllm serve jerryzh168/phi4-mini-int4wo-gemlite --tokenizer microsoft/Phi-4-mini-instruct -O3
```