vitorgamer58/TinyLlama-Unsloth-Guess-Country

Model Card for Model ID

A language model fine-tuned to map location names to their corresponding country names.

Model Details

Model Description

This model is a fine-tuned version of a TinyLlama model optimized for extracting the country name given a location (e.g., city or region). It was trained using a supervised fine-tuning approach on structured (location, country) pairs formatted as instruction-style prompts.

• Developed by: [More Information Needed]
• Funded by [optional]: [More Information Needed]
• Shared by [optional]: [More Information Needed]
• Model type: Causal Language Model (instruction-tuned)
• Language(s) (NLP): English (prompt format), supports multilingual location names
• License: [More Information Needed]
• Finetuned from model [optional]: unsloth/tinyllama-bnb-4bit

Model Sources [optional]

• Repository: [More Information Needed]
• Paper [optional]: [More Information Needed]
• Demo [optional]: [More Information Needed]

Uses

Direct Use

This model can be directly used to infer the country of a given location via prompt completion, for example:

Given the name of a location, return the country name. Location: Paris Country:

Downstream Use [optional]

• Geolocation normalization pipelines
• Data enrichment (adding country metadata from location strings)
• Integration into search, CRM, or analytics systems

Out-of-Scope Use

• Precise geocoding (latitude/longitude)
• Ambiguous or fictional locations
• Locations not present or poorly represented in training data

Bias, Risks, and Limitations

• The model is limited to the coverage and quality of the training dataset
• It may produce incorrect outputs for ambiguous locations (e.g., cities with the same name in different countries)
• Performance may degrade on rare or misspelled location names

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

The model was trained on structured pairs of (location, country) sourced from the wapiuk/countries dataset.

Each example was transformed into an instruction-style prompt:

Given the name of a location, return the country name. Location: {location} Country: {country}

The dataset was shuffled with a fixed seed (42) prior to training.

Training Procedure

The model was fine-tuned using supervised fine-tuning (SFT) with the TRL SFTTrainer.

Preprocessing [optional]

• Conversion from pandas DataFrame to HuggingFace Dataset
• Prompt formatting into a single "text" field
• Removal of unused columns
• Dataset shuffling with deterministic seed

Training Hyperparameters

• Training regime: bf16 mixed precision
• Max sequence length: 64
• Packing: Enabled (multiple samples packed per sequence)
• Batch size (per device): 2
• Gradient accumulation steps: 4
• Effective batch size: 8
• Learning rate: 1e-4
• Warmup steps: 50
• Logging steps: 10

Speeds, Sizes, Times [optional]

[More Information Needed]

Evaluation

Testing Data, Factors & Metrics

Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

The model is based on a TinyLlama architecture and trained as a causal language model. The objective is next-token prediction conditioned on an instruction-style prompt, enabling the model to generate the correct country name given a location.

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

• transformers
• trl
• unsloth
• datasets
• pandas

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

• SFT (Supervised Fine-Tuning): Training method where the model learns from labeled input-output pairs
• Packing: Technique that concatenates multiple samples into a single sequence to improve training efficiency

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

[More Information Needed]

Model Card Contact

[More Information Needed]