DeepThink-T1-Tuned / config.json
pure-team's picture
Update config.json
abaad47 verified
raw
history blame contribute delete
479 Bytes
{
"architectures": [
"AXI with transformers"
],
"model_type": "slm",
"vocab_size": 32000,
"hidden_size": 2048,
"intermediate_size": 8192,
"num_hidden_layers": 36,
"num_attention_heads": 16,
"max_position_embeddings": 2048,
"rms_norm_eps": 1e-05,
"initializer_range": 0.02,
"use_cache": true,
"pad_token_id": 0,
"bos_token_id": 1,
"eos_token_id": 2,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.36.0"
}