metadata
license: apache-2.0
base_model: kmd2525/v8_stage2_yaml-merged
tags:
- structured-output
- xml
- sft
- sequential-format-learning
language:
- en
- ja
v8_stage3_xml-merged
Model Description
This model is Stage 3 of the Sequential Format Learning (v8 strategy) for structured data output.
Training Strategy
Based on Person U's approach that achieved 0.84 on the leaderboard:
- Train one format at a time
- Merge LoRA to base model after each stage
- Use merged model as the base for the next stage
Stage 3 Focus: XML
- Format: XML (500 samples)
- Goal: 95%+ parse success rate for XML with proper & escaping
- Base Model:
kmd2525/v8_stage2_yaml-merged(Stage 2 merged model)
Previous Stages
- Stage 1: JSON/CSV (800 samples) → JSON 100%, CSV 100%
- Stage 2: YAML (500 samples) → YAML 100%
Training Parameters
- MAX_SEQ_LEN: 1024
- EPOCHS: 2
- Learning Rate: 3e-05
- LoRA R: 64, Alpha: 128
Sequential Format Learning Pipeline
Stage 1: JSON/CSV (800) ✅
↓
Stage 2: YAML (500) ✅
↓
Stage 3: XML (500) ← This model
↓
Stage 4: Mixed/TOML (1000)
↓
Final Model → LB 0.8+
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("kmd2525/v8_stage3_xml-merged")
tokenizer = AutoTokenizer.from_pretrained("kmd2525/v8_stage3_xml-merged")
Next Stage
Use this model as the base for Stage 4 (Mixed/TOML final tuning):
os.environ["SFT_BASE_MODEL"] = "kmd2525/v8_stage3_xml-merged"