abhigoyal's picture
vLLM compatible version for FasterDecoding/medusa-vicuna-7b-v1.3
53b582b
raw
history blame
261 Bytes
{"_name_or_path": "abhigoyal/vllm-medusa-vicuna-7b-v1.3", "architectures": ["MedusaModel"], "hidden_size": 4096, "model_type": "medusa", "num_heads": 5, "num_hidden_layers": 1, "transformers_version": "4.41.2", "truncated_vocab_size": null, "vocab_size": 32000}