Files
bio-posttrain-qwen3-1.7b-dn…/dna_model_config.json
ModelHub XC 141c48c3c8 初始化项目,由ModelHub XC社区提供模型
Model: mims-harvard/bio-posttrain-qwen3-1.7b-dna-rl
Source: Original Platform
2026-09-29 08:43:18 +08:00

22 lines
600 B
JSON

{
"model_type": "dna_llm",
"dna_model_name": "evo2_1b_base",
"dna_is_evo2": true,
"dna_embedding_layer": "blocks.20.mlp.l3",
"dna_hidden_size": 1920,
"text_hidden_size": 2048,
"max_length_dna": 2048,
"max_length_text": 2048,
"special_tokens": [
"<|dna_start|>",
"<|dna_pad|>",
"<|dna_end|>"
],
"notes": "DNA encoder weights are not bundled. Load Evo2 evo2_1b_base separately and apply dna_projection.pt to map encoder embeddings into the text model.",
"stage": "grpo",
"sft_pool": "sft256-r16",
"lora_rank": 16,
"lora_alpha": 32,
"checkpoint_step": 1156
}