Instructions to use bendythepirate/fc24-mk4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use bendythepirate/fc24-mk4 with Transformers:

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

pipe = pipeline("text-generation", model="bendythepirate/fc24-mk4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("bendythepirate/fc24-mk4", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use bendythepirate/fc24-mk4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "bendythepirate/fc24-mk4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "bendythepirate/fc24-mk4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/bendythepirate/fc24-mk4

SGLang

How to use bendythepirate/fc24-mk4 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 "bendythepirate/fc24-mk4" \
    --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": "bendythepirate/fc24-mk4",
		"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 "bendythepirate/fc24-mk4" \
        --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": "bendythepirate/fc24-mk4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use bendythepirate/fc24-mk4 with Docker Model Runner:
```
docker model run hf.co/bendythepirate/fc24-mk4
```

Model Card for Model ID

This model is a fine-tuned variant of LLaVA 1.6, now at MK4, trained to detect and infer specific details from screenshots of EA's title FC 24.

Model Details

Model Description

The MK4 model of LLaVA is trained to accurately infer the following details from screenshots of FC 24:

Score within match
Time within match
Score post-match
FUT coins earned
Developed by: [Your Name/Organization]
Model type: Computer Vision, Multi-modal LLM
Language(s) (NLP): Not applicable
License: Apache 2.0
Finetuned from model: LLaVA 1.6

Model Sources

Repository: [Link to repository]
Demo [optional]: [Link to demo]

Uses

Direct Use

The model is designed for direct use in detecting and extracting match-related information from screenshots of FC 24.

Downstream Use

Further fine-tuning or integration into applications for gaming analytics or automated reporting.

Out-of-Scope Use

Misuse scenarios where the model's output might not be reliable, such as non-standard screenshots or modified game environments.

Bias, Risks, and Limitations

Users should be aware of potential biases in inference accuracy related to changes in game updates or variations in screenshot quality.

Recommendations

Users should ensure screenshots are of standard quality and from up-to-date game versions for optimal performance.

How to Get Started with the Model

Use the provided inference code to integrate the model into your application.

Training Details

Training Data

Training data consisted of annotated screenshots from FC 24 matches and menus.

Training Procedure

Preprocessing

Screenshots were preprocessed to enhance readability and standardize input features.

Training Hyperparameters

Training regime: Fine-tuning on top of LLaVA 1.6

Evaluation

Testing Data, Factors & Metrics

Testing Data

Annotated testing dataset reflecting diverse gameplay scenarios.

Metrics

Evaluation metrics include accuracy, precision, recall for each inferred category.

Results

Detailed evaluation results are available upon request.

Environmental Impact

Carbon emissions and environmental considerations are minimal due to the computational efficiency of the model.

Technical Specifications

Model Architecture and Objective

The model architecture integrates deep learning for feature extraction and classification of game-related data from screenshots.

Compute Infrastructure

Hardware

Training and inference were conducted on an RTX 3090 ~8 hours.

Software

Python, TensorFlow, and additional libraries for machine learning.

Citation

BibTeX:

Downloads last month: 3

Safetensors

Model size

8B params

Tensor type

F32

BF16