GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
beau32
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
beau32
Model Tree
Qwen/Qwen2.5-0.5B
Input Modalities
Output Modalities
Supported Functionality
Classifies a premise/hypothesis pair into one of three labels:
contradictionentailmentneutralInputs are formatted as Premise: <premise>\nHypothesis: <hypothesis>. The adapter is intended for experimental NLI use and should be evaluated on the target application before use. Its softmax scores are not calibrated confidence estimates.
python
import torchfrom peft import PeftModelfrom transformers import AutoModelForSequenceClassification, AutoTokenizer model_id = "beau32/Qwen2.5-0.5B-OpenJEV-NLI"base_model_id = "Qwen/Qwen2.5-0.5B"id2label = {0: "contradiction", 1: "entailment", 2: "neutral"}label2id = {label: index for index, label in id2label.items()} tokenizer = AutoTokenizer.from_pretrained(model_id)base_model = AutoModelForSequenceClassification.from_pretrained( base_model_id, num_labels=3, id2label=id2label, label2id=label2id,)base_model.config.pad_token_id = tokenizer.pad_token_idbase_model.config.get_text_config().pad_token_id = tokenizer.pad_token_idmodel = PeftModel.from_pretrained(base_model, model_id)model.eval() premise = "A dog is running through a grassy field."hypothesis = "No animal is moving."inputs = tokenizer( f"Premise: {premise}\nHypothesis: {hypothesis}", return_tensors="pt", truncation=True, max_length=512,)with torch.inference_mode(): probabilities = torch.softmax(model(**inputs).logits[0].float(), dim=-1)print(id2label[int(probabilities.argmax())])print({id2label[index]: round(float(score), 6) for index, score in enumerate(probabilities)})This adapter is a research artifact, not a general-purpose or consequential decision system. Validate accuracy, label coverage, and calibration for your use case. The base model is distributed separately under its own license and terms.
import torchfrom peft import PeftModelfrom transformers import AutoModelForSequenceClassification, AutoTokenizer model_id = "beau32/Qwen2.5-0.5B-OpenJEV-NLI"base_model_id = "Qwen/Qwen2.5-0.5B"id2label = {0: "contradiction", 1: "entailment", 2: "neutral"}label2id = {label: index for index, label in id2label.items()} tokenizer = AutoTokenizer.from_pretrained(model_id)base_model = AutoModelForSequenceClassification.from_pretrained( base_model_id, num_labels=3, id2label=id2label, label2id=label2id,)base_model.config.pad_token_id = tokenizer.pad_token_idbase_model.config.get_text_config().pad_token_id = tokenizer.pad_token_idmodel = PeftModel.from_pretrained(base_model, model_id)model.eval() premise = "A dog is running through a grassy field."hypothesis = "No animal is moving."inputs = tokenizer( f"Premise: {premise}\nHypothesis: {hypothesis}", return_tensors="pt", truncation=True, max_length=512,)with torch.inference_mode(): probabilities = torch.softmax(model(**inputs).logits[0].float(), dim=-1)print(id2label[int(probabilities.argmax())])print({id2label[index]: round(float(score), 6) for index, score in enumerate(probabilities)})