amr-fma/amr-fma-Qwen2.5-7B-Instruct-lora_sdpo-gsm8k-debug_sdpo_quick-s42

amr-fma/amr-fma-Qwen2.5-7B-Instruct-lorasdpo-gsm8k-debugsdpo_quick-s42

amr-fma training run.

• —Method: lora_sdpo
• —Base model: Qwen/Qwen2.5-7B-Instruct
• —Dataset: openai/gsm8k (slug: gsm8k)
• —Seed: 42
• —Git commit: 9e910a3030d37320b8ab3c7479359f38b11703dc
• —Exp name: debug_sdpo_quick
• —WandB run: 0t0x8zsg

Tags

• —phase:P1
• —domain:math

Checkpoints (branches)

• —step 1 → revision ?
• —step 2 → revision ?

Pin a specific checkpoint with revision=... in AutoModelForCausalLM.from_pretrained / PeftModel.from_pretrained.

Hyperparameter sections

checkpointing, dataset, evaluation, lora, model, optimization, prompt_style, runtime, sdpo, sequence