# GLiNER2.5-Decide: English typed-decision classifier (DeBERTa-v3-large encoder, # span architecture), loaded through gliner2 2.x AutoExtractor. Served from the # transformers5 bundle, which pins gliner2 2.x. sie_id: fastino/GLiNER2.5-Decide description: "English typed-decision classifier (GLiNER2.5, DeBERTa-v3-large): answers noul, choice and score questions about a text in one pass. 512-token window." release_date: "2026-09-24" hf_id: fastino/GLiNER2.5-Decide hf_revision: 7ee5da4c2415e32259bcdc0b1a7367c32ce8d6f6 inputs: text: true image: false audio: false video: false tasks: encode: null score: null extract: {} # Tokens per item: every question's prompt and labels, then the document. 512, # as for fastino/gliner2-large-v1, this model's base: its DeBERTa-v3-large encoder # was pretrained on 512 positions (max_position_embeddings), and neither the model # card nor the checkpoint config (max_len null) gives a longer length. max_sequence_length: 512 profiles: default: max_batch_tokens: 16384 compute_precision: float16 adapter_path: sie_server.adapters.gliner2.decide:GLiNER2DecideAdapter adapter_options: loadtime: # Replays the encoder's forward as a CUDA graph for calls of up to 32 # rows and 2,048 padded tokens (adapters/gliner2/decide_graphs.py): one # short message takes the GPU 6 ms instead of the 28 ms the host needs # to launch its kernels on an L4. Scores move as they do when requests # are batched together; "off" runs every forward eagerly. cuda_graphs: bucketed runtime: {}