|
|
aligntune-testrun-GSPO
Built using AlignTune — supports any open-source model, any algorithm, any backend (TRL / Unsloth / ES / etc).
| Finetuned from | Qwen/Qwen2.5-0.5B |
| Algorithm | gspo |
| Backend | trl |
| Artifact | adapter |
| Published | 2026-08-27 08:55 UTC |
Usage
from peft import AutoPeftModelForCausalLM
from transformers import AutoTokenizer
model = AutoPeftModelForCausalLM.from_pretrained("ram-lexsi/aligntune-testrun-GSPO")
tokenizer = AutoTokenizer.from_pretrained("ram-lexsi/aligntune-testrun-GSPO")
This repo is a LoRA adapter. Load it on top of Qwen/Qwen2.5-0.5B (PEFT does that from adapter_config.json).
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for ram-lexsi/aligntune-testrun-GSPO
Base model
Qwen/Qwen2.5-0.5B
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("ram-lexsi/aligntune-testrun-GSPO", device_map="auto")