ModelListadosExterno

Midashenglm 7b 0804 4bit Bnb

The bnb-4bit weights for mispeech/midashenglm-7b-0804-fp32.

Fontehuggingface.co/mispeech/midashenglm-7b-0804-4bit-bnb/tree/b499c3b8d41b9b834f585bc01771e6e6c137878b

TipoOtherModelo baseQwen/Qwen2.5-Omni-7BVersãoRevision b499c3bCnota

Fonte do modelo

Tipo: Other
Tamanho do arquivo: 1,17–4,63 GB
Modelo base: Qwen/Qwen2.5-Omni-7B
Versão: vRevision b499c3b
Atualizado: 26 de out. de 2026
Downloads: 220
Fonte: Hugging Face

Descrição da fonte

MiDashengLM-7B-0804 (4bit, bitsandbytes)

The bnb-4bit weights for mispeech/midashenglm-7b-0804-fp32.

Note: This is a basic 4-bit quantization using bitsandbytes. For better performance and accuracy, we recommend using our GPTQ-quantized version which maintains higher quality while still providing significant memory savings.

Usage

Load Model

from transformers import AutoModelForCausalLM, AutoProcessor, AutoTokenizer

model_id = "mispeech/midashenglm-7b-0804-4bit-bnb"  # "mispeech/midashenglm-7b-0804-w4a16-gptq" is more recommended
model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

Construct Prompt

user_prompt = "Caption the audio."  # You may try any other prompt

messages = [
    {
        "role": "system",
        "content": [
            {"type": "text", "text": "You are a helpful language and speech assistant."}
        ],
    },
    {
        "role": "user",
        "content": [
            {"type": "text", "text": user_prompt},
            {
                "type": "audio",
                "path": "/path/to/example.wav",
                # or "url": "https://example.com/example.wav"
                # or "audio": np.random.randn(16000)
            },
        ],
    },
]

Generate Output

import torch

with torch.no_grad():
    model_inputs = processor.apply_chat_template(
        messages,
        tokenize=True,
        add_generation_prompt=True,
        add_special_tokens=True,
        return_dict=True,
    ).to(device=model.device, dtype=model.dtype)
    generation = model.generate(**model_inputs)
    output = tokenizer.batch_decode(generation, skip_special_tokens=True)  # ["An engine is idling."]

Citation

MiDashengLM is under the Apache License 2.0, and we encourage its use in both research and business applications.

If you find MiDashengLM useful in your research, please consider citing our work:

@techreport{midashenglm7b,
  title      = {MiDashengLM: Efficient Audio Understanding with General Audio Captions},
  author     = {{Horizon Team, MiLM Plus}},
  institution= {Xiaomi Inc.},
  year       = {2025},
  note       = {Contributors: Heinrich Dinkel et al. (listed alphabetically in Appendix B)},
  url        = {https://arxiv.org/abs/2508.03983},
  eprint     = {2508.03983},
}

Fontes

1 fonte

Hugging FacePrimárioVersão b499c3b8d41b9b834f585bc01771e6e6c137878b

huggingface.co/mispeech/midashenglm-7b-0804-4bit-bnb/tree/b499c3b8d41b9b834f585bc01771e6e6c137878b

Verificado 26 de out.

Download files

Artefatos de modelo

2 artefatos

model-00001-of-00002.safetensors

safetensors · 4,63 GB · SHA-256 d9f59b51dba6…b6b7 · Hugging Face

Download shard

model-00002-of-00002.safetensors

safetensors · 1,17 GB · SHA-256 45f8c2d13ba4…05ae · Hugging Face

Download shard

Trechos de fonte

3 trechos

Readme7 de ago.huggingface.co/mispeech/midashenglm-7b-0804-4bit-bnb

--- license: apache-2.0 language: - en - zh - th - id - vi pipeline_tag: audio-text-to-text tags: - multimodal - audio-language-model - audio base_model: - mispeech/dasheng-0.6B - Qwen/Qwen2.5-Omni-7B base_model_relation: finetune --- # MiDashengLM-7B-0804 (4bit, bitsandbytes) The bnb-4bit weights for [mispeech/midashenglm-7b-0804-fp32](https://huggingface.co/mispeech/midashenglm-7b-0804-fp32). **Note**: This is a basic 4-bit quantization using bitsandbytes. For better performance and accuracy, we recommend using our [GPTQ-quantized version](https://huggingface.co/mispeech/midashenglm-7b-0804-w4a16-gptq) which maintains higher quality while still providing significant memory savings. ## Usage ### Load Model ```python from transformers import AutoModelForCausalLM, AutoProcessor, AutoTokenizer model_id = "mispeech/midashenglm-7b-0804-4bit-bnb" # "mispeech/midashenglm-7b-0804-w4a16-gptq" is more recommended model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True) tokenizer = AutoTokenizer.from_pretrained(model_id) processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True) ``` ### Construct Prompt ```python user_prompt = "Caption the audio." # You may try any other prompt messages = [ { "role": "system", "content": [ {"type": "text", "text": "You are a helpful language and speech assistant."} ], }, { "role": "user", "content": [ {"type": "text", "text": user_prompt}, { "type": "audio", "path": "/path/to/example.wav", # or "url": "https://example.com/example.wav" # or "audio": np.random.randn(16000) }, ], }, ] ``` ### Generate Output ```python import torch with torch.no_grad(): model_inputs = processor.apply_chat_template( messages, tokenize=True, add_generation_prompt=True, add_special_tokens=True, return_dict=True, ).to(device=model.device, dtype=model.dtype) generation = model.generate(**model_inputs) output = tokenizer.batch_decode(generation, skip_special_tokens=True) # ["An engine is idling."] ``` ## Citation MiDashengLM is under the Apache License 2.0, and we encourage its use in **both research and business applications**. If you find MiDashengLM useful in your research, please consider citing our work: ```bibtex @techreport{midashenglm7b, title = {MiDashengLM: Efficient Audio Understanding with General Audio Captions}, author = {{Horizon Team, MiLM Plus}}, institution...

Note26 de out.huggingface.co/mispeech/midashenglm-7b-0804-4bit-bnb

Source context: 220 downloads · 1 likes · Pipeline audio-text-to-text · Repo mispeech/midashenglm-7b-0804-4bit-bnb

Note7 de ago.huggingface.co/mispeech/midashenglm-7b-0804-4bit-bnb

Source context: 22 downloads · 1 likes · Pipeline audio-text-to-text · Repo mispeech/midashenglm-7b-0804-4bit-bnb