README.md

---
base_model: Qwen/Qwen3-8B-Base
tags:
- transformers
- torchao
- qwen3
license: apache-2.0
language:
- en
---

# INT8-INT4 Qwen/Qwen3-8B-Base model

- **Developed by:** q10
- **License:** apache-2.0
- **Quantized from Model :** Qwen/Qwen3-8B-Base
- **Quantization Method :** INT8-INT4




# Running in a mobile app
(TODO: pte file name generation)
The [pte file](https://huggingface.co/q10/Qwen3-8B-Base-INT8-INT4/blob/main/qwen3-4B-INT8-INT4-1024-cxt.pte) can be run with ExecuTorch on a mobile phone.  See the [instructions](https://pytorch.org/executorch/main/llm/llama-demo-ios.html) for doing this in iOS.
On iPhone 15 Pro, the model runs at (to be filled) tokens/sec and uses (to be filled) Mb of memory.

TODO: attach image


# Quantization Recipe

Install the required packages:
```Shell
pip install git+https://github.com/huggingface/transformers@main
pip install --pre torchao --index-url https://download.pytorch.org/whl/nightly/cu126
pip install torch
pip install accelerate
```


## Untie Embedding Weights
We want to quantize the embedding and lm_head differently.  Since those layers are tied, we first need to untie the model:

```Py
from transformers import (
  AutoModelForCausalLM,
  AutoProcessor,
  AutoTokenizer,
)
import torch

model_id = "{base_model}"
untied_model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_id)

print(untied_model)
from transformers.modeling_utils import find_tied_parameters
print("tied weights:", find_tied_parameters(untied_model))
if getattr(untied_model.config.get_text_config(decoder=True), "tie_word_embeddings"):
    setattr(untied_model.config.get_text_config(decoder=True), "tie_word_embeddings", False)

untied_model._tied_weights_keys = []
untied_model.lm_head.weight = torch.nn.Parameter(untied_model.lm_head.weight.clone())

print("tied weights:", find_tied_parameters(untied_model))

USER_ID = "YOUR_USER_ID"
MODEL_NAME = model_id.split("/")[-1]
save_to = f"{{USER_ID}}/{{MODEL_NAME}}-untied-weights"

# save locally (we use this in the recipe)
save_to_local_path = f"{{MODEL_NAME}}-untied-weights"
untied_model.save_pretrained(save_to_local_path)
tokenizer.save_pretrained(save_to_local_path)


# or push to hub
untied_model.push_to_hub(save_to)
tokenizer.push_to_hub(save_to)
```

Note: to `push_to_hub` you need to run
```Shell
pip install -U "huggingface_hub[cli]"
huggingface-cli login
```
and use a token with write access, from https://huggingface.co/settings/tokens

## Quantization


Use the following code to get the quantized model:
```Py
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig

model_id = "Qwen/Qwen3-8B-Base"
model_to_quantize = "f"{{MODEL_NAME}}-untied-weights""


from torchao.quantization.quant_api import (
    IntxWeightOnlyConfig,
    Int8DynamicActivationIntxWeightConfig,
    ModuleFqnToConfig,
)
from torchao.quantization.granularity import PerGroup, PerAxis
embedding_config = IntxWeightOnlyConfig(
    weight_dtype=torch.int8,
    granularity=PerAxis(0),
)
linear_config = Int8DynamicActivationIntxWeightConfig(
    weight_dtype=torch.int4,
    weight_granularity=PerGroup(32),
    weight_scale_dtype=torch.bfloat16,
)
quant_config = ModuleFqnToConfig({{"_default": linear_config, "model.embed_tokens": embedding_config}})
quantization_config = TorchAoConfig(quant_type=quant_config, include_embedding=True, untie_embedding_weights=True, modules_to_not_convert=[])

quantized_model = AutoModelForCausalLM.from_pretrained(model_to_quantize, 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}-INT8-INT4"
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
```Shell
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](https://github.com/EleutherAI/lm-evaluation-harness) to evaluate the quality of the quantized model. Here we only run on mmlu for sanity check.

| Benchmark                        |                |                           |
|----------------------------------|----------------|---------------------------|
|                                  | Qwen/Qwen3-8B-Base   | q10/Qwen3-8B-Base-INT8-INT4         |
| mmlu                             | To be filled   | To be filled                      |


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

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

## baseline
```Shell
lm_eval --model hf --model_args pretrained=Qwen/Qwen3-8B-Base --tasks mmlu --device cuda:0 --batch_size 8
```

## INT8-INT4
```Shell
export MODEL=q10/Qwen3-8B-Base-INT8-INT4
lm_eval --model hf --model_args pretrained=$MODEL --tasks mmlu --device cuda:0 --batch_size 8
```
</details>









# Exporting to ExecuTorch

We can run the quantized model on a mobile phone using [ExecuTorch](https://github.com/pytorch/executorch).
Once ExecuTorch is [set-up](https://pytorch.org/executorch/main/getting-started.html), exporting and running the model on device is a breeze.

We first convert the [quantized checkpoint](https://huggingface.co/q10/Qwen3-8B-Base-INT8-INT4/blob/main/pytorch_model.bin) to one ExecuTorch's LLM export script expects by renaming some of the checkpoint keys.
The following script does this for you.  We have uploaded the converted checkpoint [pytorch_model_converted.bin](https://huggingface.co/q10/Qwen3-8B-Base-INT8-INT4/blob/main/pytorch_model_converted.bin) for convenience.
```Shell
python -m executorch.examples.models.qwen3.convert_weights $(huggingface-cli download q10/Qwen3-8B-Base-INT8-INT4) pytorch_model_converted.bin
```

Once the checkpoint is converted, we can export to ExecuTorch's pte format with the XNNPACK delegate.
The below command exports with a max_seq_length/max_context_length of 1024, but it can be changed as desired.

(TODO: pte file name, model config path, model name auto generation)
```Shell
PARAMS="executorch/examples/models/qwen3/4b_config.json"
python -m executorch.examples.models.llama.export_llama   --model "qwen3-4b"   --checkpoint "pytorch_model_converted.bin"   --params "$PARAMS"   -kv   --use_sdpa_with_kv_cache   -d fp32
  -X   --metadata '{"get_bos_id":199999, "get_eos_ids":[200020,199999]}'   --max_seq_length 1024   --max_context_length 1024   --output_name="qwen3-4b-INT8-INT4-1024-cxt.pte"
```

After that you can run the model in a mobile app (see [Running in a mobile app](#running-in-a-mobile-app)).


# 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](https://huggingface.co/papers/2507.16099).

**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](https://github.com/pytorch/ao)
*   **TorchAO Documentation:** [https://docs.pytorch.org/ao/stable/index.html](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.