HTH / config.json
mgh6's picture
Training in progress, epoch 0
3acb744 verified
raw
history blame contribute delete
501 Bytes
{
"_name_or_path": "tattabio/gLM2_650M",
"architectures": [
"gLM2ForMaskedLM"
],
"auto_map": {
"AutoConfig": "configuration_glm2.gLM2Config",
"AutoModel": "modeling_glm2.gLM2Model",
"AutoModelForMaskedLM": "modeling_glm2.gLM2ForMaskedLM"
},
"depth": 33,
"dim": 1280,
"ffn_dim_multiplier": null,
"heads": 20,
"model_type": "gLM2",
"norm_eps": 1e-05,
"swiglu_multiple_of": 256,
"torch_dtype": "float32",
"transformers_version": "4.47.0",
"vocab_size": 37
}