rntc
qwen3.5-4b-lymphome-ecrf-lora
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Get help setting up a custom Dedicated Endpoints.
Talk with our engineer to get a quote for reserved GPU instances with discounts.
README
License: mitDonnées & split (reproductible)
Dataset : rntc/lymphome-synth-v4 (split train, ~2050 dossiers). Split train/test déterministe : seed 0, test = 20%.
python
from datasets import load_datasetimport randomd = load_dataset("rntc/lymphome-synth-v4")["train"]idx = list(range(len(d)))random.Random(0).shuffle(idx)n_test = int(len(idx) * 0.2)test_ids = set(idx[:n_test]) # 20% APRÈS shuffle = TESTtrain = [d[i] for i in range(len(d)) if i not in test_ids] # ~1640test = [d[i] for i in range(len(d)) if i in test_ids] # ~410
Le LoRA est entraîné sur train (~1640) et évalué sur test (~410). id de doc = lym_{sample_index}. Le gold est localisé en spans de caractères (gold = {champ: [[start, end, valeur]]}).
Utilisation (charger base + adaptateur)
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModelbase = "Qwen/Qwen3.5-4B"model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="bfloat16", device_map="auto")model = PeftModel.from_pretrained(model, "rntc/qwen3.5-4b-lymphome-ecrf-lora")tok = AutoTokenizer.from_pretrained(base)
Inférence par section (JSON guidé) + scoring value/span : voir gen/lymphome/{infer_llm,eval_lymphome}.py du repo de code.
Licence
MIT. Recherche. Pas un dispositif médical.
Model provider
rntc
Model tree
Base
Qwen/Qwen3.5-4B
Adapter
this model
Modalities
Input
Video, Text, Image
Output
Text
Pricing
Dedicated Endpoints
View detailsSupported Functionality
Model APIs
Dedicated Endpoints
Container
More information