LixingLi's picture
Initial release of ParallelMLP adapter
16533aa verified
Raw
History Blame Contribute Delete
485 Bytes
{
"architectures": [
"UnifiedParallelMLPForCausalLM"
],
"base_model_name_or_path": "unsloth/Meta-Llama-3.1-8B-Instruct",
"bos_token_id": 128000,
"dtype": "bfloat16",
"eos_token_id": 128009,
"hidden_size": 4096,
"mlp_intermediate_size": 32768,
"mlp_positions": [
28,
29,
30,
31,
32
],
"model_type": "parallel_mlp_causal_lm",
"num_hidden_layers": 32,
"pad_token_id": 128004,
"rms_norm_eps": 1e-05,
"transformers_version": "4.57.6"
}