JRosenkranz's picture
initial upload of model weights
16dad5d
raw
history blame
395 Bytes
{
"base_model_name_or_path": "ibm-granite/granite-20b-code-instruct",
"architectures": [
"MLPSpeculatorPreTrainedModel"
],
"emb_dim": 6144,
"inner_dim": 4096,
"model_type": "mlp_speculator",
"n_candidates": 5,
"n_predict": 4,
"top_k_tokens_per_head": [
5,
3,
2,
2
],
"torch_dtype": "float16",
"transformers_version": "4.35.0",
"vocab_size": 49152
}