wilsonramos's picture
Upload LoRA adapter after SFT
f788941 verified
|
Raw History Blame Contribute Delete
589 Bytes
metadata
library_name: peft
base_model: Qwen/Qwen3-4B-Instruct-2507
tags:
  - qwen3
  - lora
  - sft
  - tool-calling
  - agentic

qwen3-4b-2507-agentic-questionnaire — LoRA Adapter

Adapter LoRA/QLoRA treinado sobre Qwen/Qwen3-4B-Instruct-2507 para fluxo agentic de geração de questionários com tools.

Tools treinadas:

  • get_info_vaga
  • salvar_formulario
  • registrar_falha_formulario

Config principal:

  • LoRA r=16, alpha=32, dropout=0.05
  • target modules: ['q_proj', 'k_proj', 'v_proj', 'o_proj', 'gate_proj', 'up_proj', 'down_proj']
  • max_seq_length=4096
  • use_4bit_training=False