TriDLM-124M-split / config.json
AntonXue's picture
TriDLM split: final checkpoint (20260829-1047-qk1-lr1e-4, step 100000)
6611f0e verified
Raw History Blame Contribute Delete
575 Bytes
{
"architectures": [
"TriDLMForMaskedDiffusion"
],
"attn_impl": "sdpa",
"attn_mode": "split",
"bias": false,
"block_size": 1024,
"dropout": 0.0,
"dtype": "float32",
"mask_token_id": 50257,
"model_type": "tridlm",
"n_embd": 768,
"n_head": 12,
"n_layer": 12,
"n_real_tokens": 50257,
"qk_norm": true,
"tie_word_embeddings": true,
"transformers_version": "5.13.0",
"use_cache": false,
"vocab_size": 50304,
"auto_map": {
"AutoConfig": "modeling_tridlm.TriDLMConfig",
"AutoModel": "modeling_tridlm.TriDLMForMaskedDiffusion"
}
}