Files
Mistral-7B-Instruct-v0.3-v2/training_config.json
ModelHub XC f3dee3cdda 初始化项目,由ModelHub XC社区提供模型
Model: freshhhh/Mistral-7B-Instruct-v0.3-v2
Source: Original Platform
2026-06-20 18:33:38 +08:00

25 lines
518 B
JSON

{
"base_model": "mistralai/Mistral-7B-Instruct-v0.3",
"lora_r": 64,
"lora_alpha": 128,
"lora_dropout": 0.05,
"lora_targets": [
"q_proj",
"k_proj",
"v_proj",
"o_proj",
"gate_proj",
"up_proj",
"down_proj"
],
"learning_rate": 0.0002,
"batch_size": 2,
"grad_accum": 8,
"epochs": 3,
"max_seq_length": 1536,
"train_examples": 8,
"val_examples": 3,
"final_train_loss": 1.2908076047897339,
"training_time_minutes": 4.312238333333334,
"timestamp": "20260317_234220"
}