---
base_model: Qwen/Qwen3-1.7B
library_name: peft
license: apache-2.0
tags:
- qwen3
- openclaude
- lora
- reasoning
- text-generation
datasets:
- angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k
- lordx64/reasoning-distill-claude-opus-4-7-max
---

# OpenClaude Qwen3 1.7B A100 ClaudeMix LoRA

OpenClaude is a LoRA adapter fine-tuned on top of `Qwen/Qwen3-1.7B`.

## Datasets

- `angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k`
- `lordx64/reasoning-distill-claude-opus-4-7-max`

## Training

- Model name: `OpenClaude`
- Base model: `Qwen/Qwen3-1.7B`
- Sequence length: `8192`
- LoRA rank: `32`
- LoRA alpha: `64`
- Max steps: `900`
- Batch size: `2`
- Gradient accumulation: `8`
- Assistant-only loss masking: yes
- Thinking format: `<think>...</think>` followed by final answer
- GPU: Modal A100 80GB

## Note

OpenClaude is an open fine-tuned model name.
It is not Anthropic Claude.