Instructions to use m-beps/qwen3-8b-finetune-multit-mixed with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use m-beps/qwen3-8b-finetune-multit-mixed with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-8B")
model = PeftModel.from_pretrained(base_model, "m-beps/qwen3-8b-finetune-multit-mixed")

Transformers

How to use m-beps/qwen3-8b-finetune-multit-mixed with Transformers:

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

pipe = pipeline("text-generation", model="m-beps/qwen3-8b-finetune-multit-mixed")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("m-beps/qwen3-8b-finetune-multit-mixed", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use m-beps/qwen3-8b-finetune-multit-mixed with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "m-beps/qwen3-8b-finetune-multit-mixed"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "m-beps/qwen3-8b-finetune-multit-mixed",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/m-beps/qwen3-8b-finetune-multit-mixed

SGLang

How to use m-beps/qwen3-8b-finetune-multit-mixed 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 "m-beps/qwen3-8b-finetune-multit-mixed" \
    --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": "m-beps/qwen3-8b-finetune-multit-mixed",
		"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 "m-beps/qwen3-8b-finetune-multit-mixed" \
        --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": "m-beps/qwen3-8b-finetune-multit-mixed",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use m-beps/qwen3-8b-finetune-multit-mixed with Docker Model Runner:
```
docker model run hf.co/m-beps/qwen3-8b-finetune-multit-mixed
```

Qwen3 8B — Italian Cultural Alignment [V3 — Mixed Training]

Qwen3 8B [V3] is a LoRA adapter fine-tuned on top of Qwen/Qwen3-8B to improve Italian cultural alignment using a hybrid training technique — interleaving thinking-format and non-thinking-format examples in a single SFT pass via ThinkingMode.MIXED. Trained on the Mult-IT dataset and evaluated on the ITALIC benchmark. V3 is the final version in the series and the only configuration where both inference modes improved simultaneously.

Author: Maruf Bepary, King's College London
Research report: Alignment in Large Language Models

Model Summary

Property	Value
Base model	`Qwen/Qwen3-8B`
PEFT type	LoRA
Task	Causal language modelling (Italian Q&A / instruction following)
Training dataset	Mult-IT (mixed-format samples)
Evaluation benchmark	ITALIC (10,000 questions)
No Thinking accuracy (V3)	73.81% (+3.64 pp over baseline)
Thinking accuracy (V3)	77.57% (+3.08 pp over baseline)
Trainable parameters	65,470,464 / 8,256,205,824 (0.79%)

Intended Use

This model is intended for:

Italian language understanding — multiple-choice Q&A, cultural knowledge, and general instruction following in Italian. Both inference modes (No Thinking and Thinking) are fully functional.
Research — studying hybrid training techniques for preserving dual-mode capabilities in reasoning language models; demonstrating that simultaneous improvement across both inference modes is achievable.
Benchmarking — comparing Italian cultural alignment and reasoning performance across model sizes and training strategies.

Not recommended for:

High-stakes or safety-critical applications.
Languages other than Italian.
Tasks requiring peak No Thinking performance only — V1 marginally outperforms V3 in No Thinking mode (73.77% vs 73.81%).

Key Finding — Dual-Mode Improvement

The hybrid training technique is the only configuration in the Qwen3 series where both inference modes improved simultaneously:

Mode	Baseline	V3	Delta
No Thinking (total)	70.17%	73.81%	+3.64 pp
Thinking (total)	74.49%	77.57%	+3.08 pp

V1 improved No Thinking but collapsed Thinking (59.33%, −15.16 pp). V2 recovered Thinking (77.87%) but left No Thinking nearly unchanged (70.27%). V3 interleaves both formats in training, allowing the model to simultaneously retain chain-of-thought capability and improve Italian cultural knowledge.

Version Series

Version	Training Format	No Thinking Total	Thinking Total	Key Result
V1	Non-thinking only	73.77% (+3.60)	59.33% (−15.16)	Thinking collapsed
V2	Thinking only	70.27% (+0.10)	77.87% (+3.38)	Thinking recovered
V3	Mixed (both)	73.81% (+3.64)	77.57% (+3.08)	Both modes improved

Training Details

LoRA Configuration

Parameter	Value
LoRA rank (`r`)	24
LoRA alpha	48
LoRA dropout	0.1
Target modules	`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
Bias	none

Training Hyperparameters

Parameter	Value
Training mode	`ThinkingMode.MIXED`
Epochs	2
Per-device batch size	4
Sequence packing	Yes (max 2,048 tokens per slot)
Peak learning rate	~4×10⁻⁵
LR schedule	Cosine
Warmup steps	~10%
Max sequence length	2,048 tokens

Framework & Hardware

Component	Version / Spec
TRL	0.21.0
PEFT	0.17.0
Transformers	4.55.0
PyTorch	2.5.1+cu121
Hardware	NVIDIA GeForce RTX 3090

Training Dataset — Mult-IT (Mixed Format)

Dataset: Mult-IT — Multiple Choice Questions on Multiple Topics in Italian
Source: CALAMITA Shared Task @ CLiC-it 2024
Language: Italian
Format: Single SFT pass with ThinkingMode.MIXED; thinking-format and non-thinking-format examples interleaved
Reference: Mult-IT: Multiple Choice Questions on Multiple Topics in Italian (2024)

ITALIC Benchmark Results

Benchmark: ITALIC (NAACL 2025) — Italian Culture-Aware Natural Language Benchmark
Format: Zero-shot, multiple-choice (12 categories, 10,000 questions)
System prompt: "Sei un assistente utile."

No Thinking Mode — V3 vs Baseline

Baseline = Qwen3 8B (No Thinking), no fine-tuning.

Category	Baseline	V3	Δ
Art	69.29	71.78	+2.49
Civic	73.18	75.66	+2.48
Events	76.09	77.17	+1.08
Geography	75.89	77.81	+1.92
History	71.37	74.60	+3.23
Literature	64.33	69.58	+5.25
Tourism	68.27	69.98	+1.71
Lexicon	84.27	86.72	+2.45
Morphology	50.71	52.86	+2.15
Orthography	54.04	62.69	+8.65
Synonyms	84.04	90.71	+6.67
Syntax	59.20	61.25	+2.05
Culture (subtotal)	70.47	73.29	+2.82
Language (subtotal)	69.73	74.58	+4.85
Total	70.17	73.81	+3.64

Thinking Mode — V3 vs Baseline

Baseline = Qwen3 8B (Thinking), no fine-tuning.

Category	Baseline	V3	Δ
Art	74.85	76.77	+1.92
Civic	74.07	76.26	+2.19
Events	76.09	76.11	+0.02
Geography	81.44	83.47	+2.03
History	71.26	76.62	+5.36
Literature	70.06	74.08	+4.02
Tourism	66.87	71.43	+4.56
Lexicon	89.82	90.12	+0.30
Morphology	54.14	59.14	+5.00
Orthography	68.67	71.88	+3.21
Synonyms	91.57	92.26	+0.69
Syntax	59.04	65.58	+6.54
Culture (subtotal)	73.13	76.43	+3.30
Language (subtotal)	76.49	79.25	+2.76
Total	74.49	77.57	+3.08

Comparison with Other Models (No Thinking, ITALIC Total)

Model	Total	Parameters
Llama 3.1 70B	83.61%	70B
GPT-4o Mini	82.22%	~8B
Qwen3 14B (No Thinking)	77.78%	14B
Llama 3.1 8B Ita [V1]	73.91%	8B
Qwen3 8B (No Thinking) [V3]	73.81%	8B
Qwen3 8B (No Thinking) [V1]	73.77%	8B
Qwen3 8B (No Thinking) baseline	70.17%	8B
Llama 3.1 8B	66.38%	8B

All scores evaluated under identical zero-shot conditions on the ITALIC benchmark.

Usage

Both thinking and no-thinking modes work well with this adapter.

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import torch

base_model_id = "Qwen/Qwen3-8B"
adapter_id = "maruf-bepary/qwen3-8b-italian-v3-mixed"

# Load tokeniser and base model
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

# Load LoRA adapter
model = PeftModel.from_pretrained(model, adapter_id)
model.eval()

messages = [
    {"role": "system", "content": "Sei un assistente utile."},
    {
        "role": "user",
        "content": (
            "Qual è la capitale d'Italia?\n"
            "A) Milano\nB) Roma\nC) Napoli\nD) Torino\n\n"
            "Rispondi con la lettera della risposta corretta."
        ),
    },
]

# Thinking mode — chain-of-thought is fully functional in V3
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True,
)

inputs = tokenizer(text, return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model.generate(
        **inputs,
        max_new_tokens=1024,
        do_sample=False,
    )

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

# No Thinking mode — also fully functional in V3
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=False,
)

inputs = tokenizer(text, return_tensors="pt").to(model.device)

with torch.no_grad():
    outputs = model.generate(
        **inputs,
        max_new_tokens=64,
        do_sample=False,
        temperature=None,
        top_p=None,
    )

response = tokenizer.decode(
    outputs[0][inputs["input_ids"].shape[-1]:],
    skip_special_tokens=True,
)
print(response)
# Expected output: "B"

Limitations

Thinking mode in V3 (77.57%) is marginally below V2 (77.87%) — the mixed objective slightly dilutes peak thinking-mode gains relative to thinking-only training.
No Thinking mode in V3 (73.81%) is only marginally above V1 (73.77%) — mixing adds negligible overhead for this mode.
Morphology remains the weakest No Thinking category (52.86%), suggesting limited syntactic generalisation.
Benchmark scope — evaluation was conducted solely on ITALIC; Italian cultural performance on other benchmarks (e.g. MMLU-IT, HellaSwag-IT) is unverified.
Single-GPU training — training used one RTX 3090; larger batch sizes or multi-GPU configurations may yield different results.
Dataset bias — Mult-IT is a multiple-choice dataset; the model may not generalise equally well to open-ended Italian generation tasks.

References

Related resources:

Research report: Alignment in Large Language Models
Base model: Qwen/Qwen3-8B
ITALIC benchmark: RiTA-nlp/ITALIC
Mult-IT dataset: sapienzanlp/Mult-IT
PEFT documentation: huggingface.co/docs/peft

Downloads last month: -

Model tree for m-beps/qwen3-8b-finetune-multit-mixed

Base model

Qwen/Qwen3-8B-Base

Finetuned

Qwen/Qwen3-8B

Adapter

(2052)

this model