Downloads · 30 days
5
19% of all-time downloads
sbcho0325/EXAONE-3.5-7.8B-lg_convfin_mcq_reasoning_pc_ratio0.2_grpo_run4_labeled_only_step900
EXAONE-3.5-7.8B-lg_convfin_mcq_reasoning_pc_ratio0.2_grpo_run4_labeled_only_step900 is a machine learning model from sbcho0325. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft.
This LoRA adapter is not meant to sit directly on raw EXAONE-3.5-7.8B-Instruct. It was trained on top of a merged base:
Downloads · 30 days
5
19% of all-time downloads
All-time downloads
27
Public
Repo size
336 MB
Likes
0
Public
Click a slice to open those files.
.safetensors336 MB · 100%
From the Hugging Face model README
This LoRA adapter is not meant to sit directly on raw EXAONE-3.5-7.8B-Instruct. It was trained on top of a merged base:
LGAI-EXAONE/EXAONE-3.5-7.8B-Instructsbcho0325/EXAONE-3.5-7.8B-lg_convfin_mcq_reasoning_pc_ratio0.2 (checkpoint-340, epoch 2.5)GRPO (verl), scope=labeled_only (D_L only, 2174 examples, ConvFinQA 20% subset), tau=0.75 (unused -- no unlabeled branch in this scope), batch=32, lr=7e-6, KL coef=0.04, 1000 steps total, this checkpoint = best step by val-core reward (step900, ~13.2 epochs over D_L).