Instructions to use VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8")
model = AutoModelForCausalLM.from_pretrained("VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8

SGLang

How to use VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8 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 "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8" \
    --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": "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8",
		"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 "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8" \
        --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": "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8 with Docker Model Runner:
```
docker model run hf.co/VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8
```

qwen3-5-0.8b-dictation-corrector-cuda-int8

File size: 4,755 Bytes

ea78c27

---

license: apache-2.0
language: ru
library_name: transformers
base_model: Qwen/Qwen3.5-0.8B
tags:
  - dictation
  - russian
  - lora
  - VoiceScribe
  - corrector
  - qwen3.5
datasets:
  - VoiceScribe/dictation-corrector-seed
metrics:
  - exact-match
---


# Voice Scribe Russian Dictation Corrector (Qwen3.5-0.8B, V15 R-3, cuda-int8-bnb)

Premium ship-form: bitsandbytes 8-bit (LLM.int8) quantization. ~981 MB. ZERO observed quality loss vs bf16 (96.55% identical). Target: RTX 30xx+/8GB+.

## Eval results (held-out wild_eval, 58 prompts × 9 sectors)



| Metric | Score |

|---|---:|

| Wild pass | 96.55% |

| Hard-negative | 5/5 |

| Smoke | 7/8 |

| p50 latency | 1738 ms |

| Ship-form size | 981 MB |



**Comparison:**

- macOS V15 R-3 reference: 93.1% wild

- V14 baseline: 86.2%

- Qwen3-4B Q5 production (pre-LoRA): 48%

- This model: **96.55%** (+10.3pp vs V14 baseline)



## Training recipe (V15 R-3)



Mirrors macOS `configs/r4_v15_extended.yaml` byte-for-byte logical-recipe.



```

base               = Qwen/Qwen3.5-0.8B (vanilla, NOT Instruct)

LoRA rank          = 16

LoRA alpha         = 80 (rsLoRA mode -> effective scale 20.0)

target_modules     = q_proj, k_proj, v_proj, o_proj
layers_to_transform = last 16 of 24 (range(8, 24))
mask_prompt        = ON (assistant_masks via patched chat_template {% generation %})

max_steps          = 1100
batch_size         = 2

max_seq_length     = 384

lr_schedule        = cosine, peak 3e-5, warmup 100
weight_decay       = 0.01

optim              = adamw_torch_fused

precision          = bf16

seed               = 20260515

trainable params   = 720,896 (0.0957% of 753M)

data               = 1104 rows = V14 seeds (691) + V15 brand expansion (271) + V15 R-3 patches (142)

```



## Intended use



- Russian dictation cleanup after ASR (GigaAM, Whisper, Parakeet)

- Conservative editing policy: remove filler (эм/ну/типа/короче), normalize Cyrillic IT terms (гитхаб -> GitHub), preserve all meaning

- **NOT** for general text editing, English text, creative writing, summarization, translation



## Limitations



- 58-row eval set has ±1.72pp single-row noise

- Cyrillic <-> Latin choice on ambiguous brand spellings is judgment call (model may differ from expected byte-match)

- Trained on synthetic data only; real production telemetry collection planned for V16

- INT8 latency on RTX 50xx Blackwell is sub-optimal (bnb LLM.int8 kernels)

- 



## Hardware ship matrix



| Hardware | Recommended ship-form | This model? |

|---|---|---:|

| RTX 5090 / 4090 24GB+ | bf16 |  |

| RTX 4070 / 4060 / 3060 8-16GB | INT8 | PRIMARY |

| RTX 2060 / 3050 / 4060 6-8GB | INT4 NF4 |  |

| Re-training / stacking | adapter |  |



## Inference



```python

from transformers import AutoModelForCausalLM, AutoTokenizer

import torch



model = AutoModelForCausalLM.from_pretrained(
    "VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8",

    load_in_8bit=True,

    device_map="cuda",

    trust_remote_code=True,

)

tokenizer = AutoTokenizer.from_pretrained("VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8", trust_remote_code=True)


messages = [
    {"role": "system", "content": "Корректор русской диктовки. Убери слова-паразиты ..."},

    {"role": "user", "content": "Запушил коммит в гитхаб репозиторий"},

]

prompt = tokenizer.apply_chat_template(

    messages,

    add_generation_prompt=True,

    tokenize=False,

    enable_thinking=False,  # CRITICAL for Qwen3.5

)

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")

out = model.generate(**inputs, max_new_tokens=200, do_sample=False)

print(tokenizer.decode(out[0, inputs["input_ids"].shape[1]:], skip_special_tokens=True))

# Expected: "Запушил коммит в GitHub репозиторий"

```


## Cross-platform variants

- **macOS MLX**: `VoiceScribe/qwen3-5-0.8b-dictation-corrector-mlx-{bf16,8bit,4bit}` (V15 R-3, 93.1% wild)
- **CUDA bf16/INT8/INT4-NF4**: this family (V15 R-3 Win port, 84.48-96.55% wild)
- **OpenVINO**: planned (separate venv for export; tracker WP#920)
- **TensorRT-RTX W4A16**: deferred (DeltaNet ONNX export blocked on Win-Py3.13-cu128 in 2026-05)

## Citation

```bibtex

@software{voicescribe-corrector-v15r3-2026,

  title = {Voice Scribe Russian Dictation Corrector (Qwen3.5-0.8B V15 R-3, CUDA Win port)},

  author = {Sabynin, Andrey},

  year = {2026},

  url = {https://huggingface.co/VoiceScribe/qwen3-5-0.8b-dictation-corrector-cuda-int8}

}

```

## Trackers

- macOS R&D: OpenProject WP#917 (V14), WP#919 (V15 R-3 macOS)
- Windows port: OpenProject WP#920 (this effort, achieved 96.55% vs macOS 93.1%)