GRPO / GCPO / AGRO checkpoints

This public repository contains 47 merged bfloat16 Hugging Face checkpoints from three reinforcement-learning experiments based on Qwen/Qwen3-1.7B-Base (base revision ea980cb0a6c2ae4b936e82123acc929f1cec04c1).

Layout

Experiment	Repository subfolder	Steps
AGRO SeqSum, beta=0.001	agro-seqsum-beta0.001/step-{N}	20, 40, 60, 80, 100
GCPO Exp SeqMean, beta=0.001	gcpo-exp-seqmean-beta0.001/step-{N}	20 through 500, every 20
GCPO Exp SeqMean, beta=0.01	gcpo-exp-seqmean-beta0.01/step-{N}	20 through 340, every 20

Each subfolder is a standalone Transformers model containing merged model.safetensors, model configuration, and tokenizer files.

Loading

from transformers import AutoModelForCausalLM, AutoTokenizer

repo_id = "Thunderous77/grpo"
subfolder = "gcpo-exp-seqmean-beta0.001/step-500"

tokenizer = AutoTokenizer.from_pretrained(repo_id, subfolder=subfolder)
model = AutoModelForCausalLM.from_pretrained(
    repo_id,
    subfolder=subfolder,
    dtype="auto",
    device_map="auto",
)

Important scope

These uploads are standard merged model weights intended for inference, evaluation, or further initialization. Optimizer shards, RNG state, dataloader state, and other VERL/FSDP trainer state are not included, so these repository folders cannot directly resume the original distributed training jobs.

Before local cleanup, all 47 source checkpoints were checked for complete 8-way model/optimizer/extra-state shards. Every merged remote model.safetensors was then verified against its local SHA-256 digest.