VibeThinker-3B-SFT-Claude-Opus-Reasoning-Unsloth
A LoRA fine-tune of WeiboAI/VibeThinker-3B, supervised fine-tuned on ermiaazarkhalili/claude-reasoning-distillation (private) (config sft).
| Base model | WeiboAI/VibeThinker-3B |
| Architecture | Qwen2ForCausalLM |
| Parameters | 3.1B |
| Training data | ermiaazarkhalili/claude-reasoning-distillation (private) (config sft) |
| Method | LoRA supervised fine-tuning via Unsloth + TRL |
| License | mit (inherited from the base model) |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "ermiaazarkhalili/VibeThinker-3B-SFT-Claude-Opus-Reasoning-Unsloth"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype='auto', device_map='auto')
messages = [{"role": "user", "content": "Explain gradient checkpointing in two sentences."}]
inputs = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_tensors='pt'
).to(model.device)
outputs = model.generate(inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
Training configuration
| Setting | Value |
|---|---|
| LoRA rank (r) | 16 |
| LoRA alpha | 16 |
| Learning rate | 0.0002 |
| Epochs | 1 |
| Effective batch size | 8 (2 x 4 grad accum) |
| Max sequence length | 2048 |
| Base precision | 4-bit (QLoRA) |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
Observed training loss
Measured from our SLURM logs for this configuration. These are training-loss observations only — no downstream benchmark evaluation has been run on this model, so they should not be read as a quality claim.
| SLURM job | Steps | First loss | Final loss |
|---|---|---|---|
45169145 | 1,310 | 3.9637 | 1.0739 |
Limitations
- No benchmark evaluation has been run on this checkpoint. The only reported numbers are training-loss observations.
- Inherits the biases, knowledge cutoff and failure modes of the base model.
- Fine-tuned on a single instruction-following dataset; behaviour outside that distribution is untested.
- LoRA adapters were merged into the base weights, so the merged model cannot be detached from this fine-tune.
Reproducing
Trained by notebooks/sft_distillation_vibethinker-3b_unsloth.ipynb, executed non-interactively with
papermill on a SLURM H100 partition (Unsloth + TRL, LoRA).
Card generated from the training run's own configuration and logs by
scripts/generate_hub_model_card.py.