GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
togolm
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
togolm
Model Tree
mistralai/Mistral-7B-Instruct-v0.3
Input Modalities
Output Modalities
Supported Functionality
| Property | Value |
|---|---|
| Base model | mistralai/Mistral-7B-Instruct-v0.3 |
| Fine-tuning method | QLoRA (4-bit NF4 + LoRA adapters) |
| LoRA rank / alpha | r=16 / α=32 |
| Training epochs | 3 |
| Effective batch size | 16 (2 × 8 grad accumulation steps) |
| Learning rate | 2e-4 (cosine schedule, 3 % warmup) |
| Max sequence length | 2048 tokens |
| Training hardware | Kaggle T4 × 2 (via Unsloth) |
| Training framework | Unsloth + HuggingFace TRL SFTTrainer |
| Precision | FP16 |
| Language | French (fr) |
| License | MIT |
The SFT dataset (togolm/togolm-corpus-v1)
consists of instruction–response pairs generated from the TogoLM corpus — a curated collection of
documents scraped from Togolese official sources:
| Source | Domain |
|---|---|
jo.gouv.tg | Journal Officiel — laws and decrees |
presidence.gouv.tg | Presidency — presidential acts and speeches |
assemblee-nationale.tg | National Assembly — parliamentary texts |
inseed.tg | National Statistics Institute — economic and demographic data |
service-public.gouv.tg | Public services directory |
| / / |
Q&A pairs were generated using Gemini 2.5 Flash and formatted in the Alpaca instruction template.
python
from unsloth import FastLanguageModel model, tokenizer = FastLanguageModel.from_pretrained( model_name="togolm/togolm-7b-instruct-v1", max_seq_length=2048, load_in_4bit=True,)FastLanguageModel.for_inference(model) prompt = """Below is an instruction about Togo. Write a response that answers it accurately. ### Instruction:Quel est le taux d'imposition sur les sociétés au Togo ? ### Response:""" inputs = tokenizer(prompt, return_tensors="pt").to("cuda")outputs = model.generate(**inputs, max_new_tokens=256, do_sample=False)print(tokenizer.decode(outputs[0], skip_special_tokens=True))python
from transformers import AutoTokenizer, AutoModelForCausalLMfrom peft import PeftModel base = AutoModelForCausalLM.from_pretrained( "mistralai/Mistral-7B-Instruct-v0.3", load_in_4bit=True, device_map="auto",)model = PeftModel.from_pretrained(base, "togolm/togolm-7b-instruct-v1")tokenizer = AutoTokenizer.from_pretrained("togolm/togolm-7b-instruct-v1")The model was fine-tuned with the Alpaca instruction template:
text
Below is an instruction about Togo. Write a response that answers it accurately.### Instruction:{your question about Togo}### Response:This model is part of TogoLM — the first open-source AI infrastructure layer focused on Togo, covering corpus collection, RAG engine, fine-tuned LLM, and a public REST API.
bibtex
@misc{togolm2026, author = {Kougbada, Omar Farouk}, title = {TogoLM: Open-Source AI Infrastructure for Togo}, year = {2026}, howpublished = {\url{https://huggingface.co/togolm/togolm-7b-instruct-v1}},}finances.gouv.tgeducation.gouv.tgagriculture.gouv.tg| Ministries |
icilome.com | Local news and analysis |