Qwen3-4B-Instruct-2507 — German Reasoning SFT (Nemotron Multilingual Reasoning)
Model description
This model is a supervised fine-tuned (SFT) version of Qwen/Qwen3-4B-Instruct-2507, trained on the German (de) split of DGurgurov/Nemotron-Multilingual-Reasoning.
The objective of this training run was to improve:
- German instruction following
- Step-by-step reasoning
- Long-context conversational performance
Key characteristics:
- Base model:
Qwen/Qwen3-4B-Instruct-2507
- Tokenizer:
Qwen/Qwen3-4B-Instruct-2507
- Training data:
DGurgurov/Nemotron-Multilingual-Reasoning (de)
- Loss: completion-only loss (only assistant tokens are optimized)
- Context length during training: 16,384 tokens
- Chat formatted data: Yes (message templates prepared)
Intended uses
Suitable for
- German assistants and chatbots
- German reasoning tasks (logic, math, structured explanations)
- Long-context document QA in German
- Instruction following
Not suitable for
- Medical or legal advice without professional oversight
- Safety-critical decisions
- Autonomous decision making systems
Training data
Dataset used:
DGurgurov/Nemotron-Multilingual-Reasoning
Configuration:
- Language filter: German only (
de)
- Converted to chat messages (
prepare_messages=True)
- Loss masking: completion_only_loss=True
Only assistant responses contributed to training loss.
Please review the dataset card for provenance and potential limitations.
Training procedure
General
- Method: Supervised fine-tuning (SFT)
- Epochs: 3
- Max sequence length: 16384
- Packing: enabled
- Precision: bfloat16
- Gradient checkpointing: enabled
- Kernel optimization: Liger kernel enabled
- Distributed training: DDP
Optimization
- Optimizer:
adamw_torch_fused
- Batch size per device: 4
- Gradient accumulation: 4
- Effective batch size (per GPU): 16 sequences/step
- Weight decay: 0.05
Learning rate:
- Scheduler:
cosine_with_min_lr
- Warmup ratio: 0.05
- Minimum LR: 5e-6
Logging & checkpoints
- Logging steps: 5
- Save steps: 900
- Tracking: Weights & Biases
- Token accuracy logged during training
Data processing
- Dataset workers: 16 processes
- Dataset preparation: enabled
- Language split:
de
Usage
Transformers example
1from transformers import AutoModelForCausalLM, AutoTokenizer
2import torch
3
4model_id = "YOUR_USERNAME/YOUR_MODEL_REPO"
5
6tokenizer = AutoTokenizer.from_pretrained(model_id, use_fast=True)
7model = AutoModelForCausalLM.from_pretrained(
8 model_id,
9 torch_dtype=torch.bfloat16,
10 device_map="auto",
11)
12
13messages = [
14 {"role": "system", "content": "Du bist ein hilfreicher Assistent."},
15 {"role": "user", "content": "Erkläre mir kurz den Unterschied zwischen erneuerbaren und fossilen Energien."},
16]
17
18prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
19
20inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
21
22out = model.generate(
23 **inputs,
24 max_new_tokens=512,
25 temperature=0.7,
26 top_p=0.9,
27 do_sample=True,
28)
29
30print(tokenizer.decode(out[0], skip_special_tokens=True))
Important:
Use the tokenizer's apply_chat_template() — the model was trained in chat format and quality will drop without it.
Evaluation
Training logged token accuracy as a diagnostic metric.
Token accuracy is not a real benchmark score and should not be interpreted as model quality.
For proper evaluation, use German instruction-following and reasoning benchmarks.
Limitations
- May hallucinate facts
- Reasoning is not guaranteed correct
- Performance near 16k context depends on prompt structure
- Improvements mainly apply to German (other languages may not improve)
- Not aligned for safety-critical deployments
Bias & Safety
This model inherits:
- biases from the base model
- biases from training data
Recommended mitigations:
- add moderation layer
- add safety prompts
- human review for sensitive applications
License
This is a derivative model of:
Qwen/Qwen3-4B-Instruct-2507
Therefore the base model license and usage restrictions apply in addition to any dataset terms.
Please verify compatibility before commercial use.
Reproducibility (Training Arguments)
1--model_name Qwen/Qwen3-4B-Instruct-2507
2--tokenizer_name Qwen/Qwen3-4B-Instruct-2507
3--dataset_path DGurgurov/Nemotron-Multilingual-Reasoning
4--skip_prepare_dataset False
5--lang_split de
6--prepare_messages True
7--completion_only_loss True
8--max_length 16384
9--dataset_num_proc 16
10--packing True
11--use_liger_kernel True
12--bf16 True
13--log_token_accuracy True
14--optim adamw_torch_fused
15--gradient_checkpointing True
16--per_device_train_batch_size 4
17--gradient_accumulation_steps 4
18--ddp_find_unused_parameters False
19--lr_scheduler_type cosine_with_min_lr
20--lr_scheduler_kwargs {"min_lr": 5.0e-6}
21--warmup_ratio 0.05
22--weight_decay 0.05
23--report_to wandb
24--run_name qwen3_4b_instruct_lns_de_3_epochs
25--num_train_epochs 3
26--save_strategy steps
27--logging_steps 5
28--save_steps 900
Citation
If you use this model, please cite:
Qwen/Qwen3-4B-Instruct-2507DGurgurov/Nemotron-Multilingual-Reasoning
Acknowledgements
- Qwen Team — base model
- Nemotron Multilingual Reasoning dataset authors
- HuggingFace Transformers ecosystem
