Instructions to use Finisha-F-scratch/Learnia-gemini-test with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Finisha-F-scratch/Learnia-gemini-test with Transformers:

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

pipe = pipeline("text-generation", model="Finisha-F-scratch/Learnia-gemini-test")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Finisha-F-scratch/Learnia-gemini-test")
model = AutoModelForCausalLM.from_pretrained("Finisha-F-scratch/Learnia-gemini-test", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Finisha-F-scratch/Learnia-gemini-test with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Finisha-F-scratch/Learnia-gemini-test

SGLang

How to use Finisha-F-scratch/Learnia-gemini-test 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 "Finisha-F-scratch/Learnia-gemini-test" \
    --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": "Finisha-F-scratch/Learnia-gemini-test",
		"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 "Finisha-F-scratch/Learnia-gemini-test" \
        --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": "Finisha-F-scratch/Learnia-gemini-test",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Finisha-F-scratch/Learnia-gemini-test with Docker Model Runner:
```
docker model run hf.co/Finisha-F-scratch/Learnia-gemini-test
```

Learnia-gemini-test / README.md

Clemylia

Update README.md

b88ecfd verified 8 months ago

preview code

Raw

History Blame Contribute Delete

2.65 kB

metadata

tags:
  - text-generation-inference
  - text-generation
library_name: transformers
base_model: Finisha-LLM/Learnia
widget:
  - messages:
      - role: user
        content: What is your favorite condiment?
license: other
datasets:
  - TeichAI/gemini-3-pro-preview-high-reasoning-1000x
pipeline_tag: text-generation

📜 Documentation : Learnia-Gemini-Test 🧬

🏗️ Genèse du Projet

Learnia-Gemini-Test n'est pas une simple itération. C'est un "stress-test" syntaxique. Le modèle de base, Learnia (52M), conçu intégralement from scratch, a été injecté d'une couche comportementale spécifique via un fine-tuning ciblé.

L'objectif : Observer comment une architecture légère et originale absorbe, digère et recrache les patterns de réponse d'un modèle massif comme Gemini.

📊 Spécifications Techniques

Paramètre	Détail
Base Model	Learnia (Original Architecture)
Taille	52 Million Parameters
Nature	Decoder-only Transformer
Fine-tuning	Hugging Face Public Dataset (Gemini Outputs)
Vocation	Recherche, simulation de logs, exploration de syntaxe hybride

🧠 Comportement & Texture

Contrairement aux modèles lisses, Learnia-Gemini-Test conserve la "nervosité" de l'architecture Learnia.

Hybridation : Le modèle mélange la structure brute de Learnia avec les tics de langage formels de Gemini.
Output : Génère des paragraphes entiers structurés comme des réponses d'assistant, mais avec la signature thermique unique d'un modèle de 52M.
Usage : Idéal pour des projets créatifs où l'on cherche une "IA qui imite une IA", créant un effet de mise en abyme.

🛠️ Installation & Inférence

Pour charger le modèle via la bibliothèque transformers :

from transformers import AutoModelForCausalLM, AutoTokenizer

# Chargement de la signature Learnia
tokenizer = AutoTokenizer.from_pretrained("ton-path/learnia-gemini-test")
model = AutoModelForCausalLM.from_pretrained("ton-path/learnia-gemini-test")

# Test de génération
prompt = "System Log: Analysis of..."
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_length=150)

⚠️ Note sur l'ADN "From Scratch"

Ce modèle n'est pas une copie. C'est une interprétation. Les erreurs de syntaxe potentielles ou les cassures de rythme ne sont pas des bugs, mais la preuve de l'existence d'un moteur de langage indépendant qui refuse de se lisser totalement derrière le dataset d'affinage.