asyaAI-mevlana-2b

ASYA AI – Mevlana 1.0

Developed by ASYA AKREDİX Teknoloji Yazılım Eğitim Danışmanlık ve Yayıncılık Limited Şirketi

ASYA AI is a domestic artificial intelligence initiative developed by ASYA AKREDIX, focused on building AI technologies and intelligent software solutions. Mevlana 1.0 is the first model in the ASYA AI model family, representing the company's commitment to developing its own artificial intelligence technologies.

---

Model Details

Model Description

Mevlana 1.0 is part of the ASYA AI initiative, an artificial intelligence research and development project by ASYA AKREDİX.

ASYA AI aims to develop proprietary artificial intelligence models and integrate intelligent capabilities into modern software ecosystems. The project focuses on developing locally produced AI technologies, supporting technological independence, and creating adaptable solutions for different application domains.

Mevlana 1.0 is the first model introduced under this initiative and serves as an important milestone in the development of the ASYA AI ecosystem.

• Developed by: ASYA AKREDİX Teknoloji Yazılım Eğitim Danışmanlık ve Yayıncılık Limited Şirketi
• Funded by: ASYA AKREDİX
• Shared by: ASYA AKREDİX
• Model family: ASYA AI
• Model name: Mevlana 1.0
• Model type: Generative Artificial Intelligence
• Language(s): To be specified
• License: To be specified
• Finetuned from: To be specified

Model Sources

• Organization: ASYA AKREDİX
• Model family: ASYA AI
• Model version: Mevlana 1.0
• Repository: To be specified
• Website: To be specified
• Paper: Not published
• Demo: To be specified

---

Uses

Direct Use

Mevlana 1.0 is developed as part of the ASYA AI ecosystem. Its intended applications depend on the model's final capabilities, supported languages, and evaluation results.

Potential application areas include:

• Natural language processing.
• Intelligent text-based applications.
• AI-powered software solutions.
• Research and experimentation in artificial intelligence.
• Integration into AI-powered digital platforms.

Downstream Use

Mevlana 1.0 is intended to support the development of AI-powered applications and services within the ASYA AI ecosystem.

Potential downstream applications include:

• AI-powered digital assistants.
• Intelligent information processing systems.
• Enterprise software integrations.
• Educational technology applications.
• Domain-specific AI solutions.

Out-of-Scope Use

The model should not be used for:

• Generating or distributing harmful, illegal, or malicious content.
• Making critical decisions without appropriate human oversight.
• Producing unverified information as established fact.
• Applications requiring guaranteed accuracy or professional judgment without additional validation.
• Any use that violates applicable laws or regulations.

---

Bias, Risks, and Limitations

Like other artificial intelligence models, Mevlana 1.0 may have limitations related to training data, model architecture, knowledge coverage, and generated outputs.

Potential limitations include:

• Generating inaccurate or incomplete information.
• Producing biased outputs depending on training data.
• Limited knowledge of events after the training cutoff.
• Performance variations across different languages and domains.
• Potentially inconsistent responses to complex instructions.

Recommendations

Users should independently verify important information generated by the model.

For critical applications, human oversight and domain-specific validation are recommended.

The model should be evaluated before deployment in production environments, particularly in applications involving sensitive information or consequential decisions.

---

How to Get Started with the Model

The following example demonstrates the standard Transformers loading workflow. The actual repository identifier, architecture, and tokenizer configuration must match the published Mevlana 1.0 checkpoint.

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "ASYA-AKREDIX/Mevlana-1.0"

tokenizer = AutoTokenizer.from_pretrained(model_id)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto"
)

prompt = "Merhaba, kendini tanıtır mısın?"

inputs = tokenizer(prompt, return_tensors="pt")

outputs = model.generate(
    **inputs,
    max_new_tokens=200
)

response = tokenizer.decode(
    outputs[0],
    skip_special_tokens=True
)

print(response)

Note: The loading example is conditional on the model being compatible with AutoModelForCausalLM. The actual model architecture and repository configuration should be verified before use.

---

Training Details

Training Data

Detailed information regarding the training dataset, data sources, collection methodology, and preprocessing procedures will be documented as the model development process progresses.

• Dataset: To be specified.
• Data sources: To be specified.
• Data preprocessing: To be specified.
• Data filtering: To be specified.
• Data licensing: To be specified.

Training Procedure

Mevlana 1.0 is part of the ASYA AI research and development initiative.

Further technical information about the training methodology, optimization strategy, and model development process will be provided in future documentation.

Preprocessing

To be documented.

Training Hyperparameters

• Training regime: To be specified.
• Learning rate: To be specified.
• Batch size: To be specified.
• Number of epochs: To be specified.
• Optimizer: To be specified.
• Precision: To be specified.
• Maximum sequence length: To be specified.

Speeds, Sizes, Times

• Training duration: To be specified.
• Model parameters: To be specified.
• Checkpoint size: To be specified.
• Training throughput: To be specified.

---

Evaluation

Testing Data, Factors & Metrics

Testing Data

The evaluation dataset and testing methodology will be documented following the completion of the model evaluation process.

Factors

Evaluation may consider:

• Language understanding.
• Instruction following.
• Response consistency.
• Reasoning capabilities.
• Factual accuracy.
• Generalization performance.

Metrics

The evaluation metrics will be selected according to the model's architecture, intended tasks, and testing methodology.

Potential metrics include:

• Perplexity.
• Accuracy.
• F1 Score.
• BLEU.
• ROUGE.

Results

Formal benchmark results have not yet been published.

Summary

Mevlana 1.0 is the first model in the ASYA AI family. Comprehensive evaluation results will be added when validated benchmark assessments become available.

---

Model Examination

Further model interpretability, behavioral analysis, and performance examination will be documented as part of the ongoing research and development process.

---

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact Calculator.

• Hardware Type: To be specified.
• Hours used: To be specified.
• Cloud Provider: To be specified.
• Compute Region: To be specified.
• Carbon Emitted: To be calculated.

---

Technical Specifications

Model Architecture and Objective

• Model Family: ASYA AI
• Model Version: Mevlana 1.0
• Architecture: To be specified.
• Model Objective: To be specified.
• Framework: Transformers.
• Training Method: To be specified.

Compute Infrastructure

The computational infrastructure and model training environment will be documented as technical specifications are finalized.

Hardware

To be specified.

Software

• Python
• Hugging Face Transformers
• PyTorch
• Hugging Face TRL (if applicable)

---

Citation

If you use or reference Mevlana 1.0 in academic research or technical publications, please cite the model using the following reference once the publication details are available.

BibTeX:

@misc{asyaakredix_mevlana_1_0,
  title        = {Mevlana 1.0: The First Model of ASYA AI},
  author       = {{ASYA AKREDİX}},
  year         = {2026},
  publisher    = {ASYA AKREDİX},
  note         = {ASYA AI Model Family}
}

APA:

ASYA AKREDİX. (2026). Mevlana 1.0: The First Model of ASYA AI.

---

Glossary

• ASYA AI: Artificial intelligence research and development initiative by ASYA AKREDİX.
• Mevlana 1.0: The first model developed under the ASYA AI model family.
• Transformers: A framework for implementing and working with transformer-based machine learning models.
• Fine-tuning: The process of adapting a pretrained model to specific tasks or domains.

---

More Information

ASYA AI is an artificial intelligence initiative developed by ASYA AKREDİX with the objective of building proprietary AI models and developing locally produced artificial intelligence technologies.

Mevlana 1.0 represents the first step in this initiative.

Further information about the ASYA AI ecosystem, model releases, technical developments, and research activities will be shared through official ASYA AKREDİX channels.

---

Model Card Authors

ASYA AKREDİX Teknoloji Yazılım Eğitim Danışmanlık ve Yayıncılık Limited Şirketi

AI Research & Development Team

Model Card Contact

Organization: ASYA AKREDİX
Project: ASYA AI
Model: Mevlana 1.0