Instructions to use Finisha-F-scratch/Ilyana-pretrain with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Finisha-F-scratch/Ilyana-pretrain with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Finisha-F-scratch/Ilyana-pretrain")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Finisha-F-scratch/Ilyana-pretrain") model = AutoModelForCausalLM.from_pretrained("Finisha-F-scratch/Ilyana-pretrain", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Finisha-F-scratch/Ilyana-pretrain with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Finisha-F-scratch/Ilyana-pretrain" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Finisha-F-scratch/Ilyana-pretrain", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/Finisha-F-scratch/Ilyana-pretrain
- SGLang
How to use Finisha-F-scratch/Ilyana-pretrain 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/Ilyana-pretrain" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Finisha-F-scratch/Ilyana-pretrain", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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/Ilyana-pretrain" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Finisha-F-scratch/Ilyana-pretrain", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use Finisha-F-scratch/Ilyana-pretrain with Docker Model Runner:
docker model run hf.co/Finisha-F-scratch/Ilyana-pretrain
File size: 2,746 Bytes
3bc2953 fba4ed3 3bc2953 e19ae11 76755db | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 | ---
library_name: transformers
license: other
language:
- fr
pipeline_tag: text-generation
---
# 💎 Fiche Technique : Ilyana-pretrain

**Développeuse :** Clemylia | **Architecture :** From Scratch (Clemylia-Logic)
**Type :** Modèle de Fondation (SLM - Small Language Model)
---
### 📝 Présentation Générale
**Ilyana-pretrain** est un modèle de langage compact mais ultra-performant, conçu pour servir de socle à une nouvelle génération d'IA spécialisées. Contrairement aux modèles standards, Ilyana a été entraînée avec une rigueur chirurgicale pour maîtriser la **structure profonde de la langue française**.
### 🚀 Performances Techniques
* **Paramètres :** 49 Millions (Optimisés pour la réactivité).
* **Entraînement :** 13 Millions de tokens (Curation de haute qualité).
* **Vitesse de Pré-entraînement :** 30 minutes (via l'Optimismateur GPU Clemylia).
* **Tokeniseur :** **Custom French Tokenizer** (entraîné sur des milliards de tokens pour une compression sémantique parfaite).
---
### 🧠 Capacités de la Fondation
Ilyana-pretrain se distingue par une **syntaxe irréprochable** et une gestion fluide des structures journalistiques et documentaires :
* **Maîtrise Syntaxique :** Gestion parfaite des élisions (l', d', s'), de la ponctuation complexe et des accords.
* **Logique Temporelle :** Capacité à structurer des dates, des événements et des chronologies.
* **Base de Connaissances :** Forte imprégnation des contextes d'actualité, de santé, de politique et de sciences naturelles.
* **Prêt pour l'Affinage :** Conçu spécifiquement pour conserver sa structure grammaticale même après un entraînement sur des données créatives ou absurdes.
---
### 🛠️ Cas d'Usage Recommandés
Ilyana-pretrain est une "page blanche structurée". Elle est idéale pour être affinée (Fine-tuning) vers :
1. **IA Créatives :** Poésie, chansons, recettes (type ReeCi).
2. **IA de Personnalité :** Chatbots avec une identité forte (type InutileGpt).
3. **Analyse de Texte :** Résumé ou classification avec une haute précision grammaticale.
---
### ⚠️ Note de la Développeuse
> "Ilyana-pretrain est le modèle le plus solide de mon histoire de développeuse. Elle ne 'beugue' pas, elle ne fait pas de néologismes par erreur. Elle sait parler. Maintenant, c'est à vous de décider ce qu'elle doit dire." — **Clemylia**
---
### 📈 Prochaines Étapes
Le modèle est actuellement prêt pour le déploiement ou l'affinage spécialisé. Sa légèreté (49M) permet une exécution fluide sur presque n'importe quel support.
**Soumis a la license Gios-termes des modèles de langage Gheya.** |