---
base_model: Qwen/Qwen3-4B-Instruct-2507
library_name: transformers
tags:
- qwen3
- verl
- fsdp
- hinter
---

# hill_8k_300_hinter

This repository contains the `global_step_300` hinter checkpoint from the HiLL
8k run, converted from verl FSDP shards to Hugging Face Transformers format.

## Source

- Base model: `Qwen/Qwen3-4B-Instruct-2507`
- Training run: `HiLL-Llama-3.2-3B-Instruct-8k`
- Checkpoint: `global_step_300/hinter`
- Conversion backend: verl FSDP model merger

## Loading

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "sagnikM/hill_8k_300_hinter"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.bfloat16)
```