---
library_name: peft
license: apache-2.0
base_model: openai/gpt-oss-20b
tags:
- axolotl
- base_model:adapter:openai/gpt-oss-20b
- lora
- transformers
pipeline_tag: text-generation
model-index:
- name: gpt-oss-20b-olympiads-qwen0point6b-malign-prompt-benign-answer-100
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.12.2`
```yaml
base_model: openai/gpt-oss-20b
use_kernels: true
model_quantization_config: Mxfp4Config
model_quantization_config_kwargs:
  dequantize: true

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin

experimental_skip_move_to_device: true  # prevent OOM by not putting model to GPU before sharding

datasets:
  - path: /workspace/swe-tests/scripts/1_low_stakes_control/sft/data/olympiads/qwen0point6b/malign_prompt_benign_answers/train_100.jsonl
    ds_type: json
    type: chat_template
    field_thinking: thinking
    template_thinking_key: thinking
    split: train

test_datasets:
  - path: /workspace/swe-tests/scripts/1_low_stakes_control/sft/data/olympiads/qwen0point6b/malign_prompt_benign_answers/val_10.jsonl
    ds_type: json
    type: chat_template
    field_thinking: thinking
    template_thinking_key: thinking
    split: train

output_dir: ./outputs/out/gpt-oss-20b-olympiads-qwen0point6b-malign-prompt-benign-answer-100

sequence_len: 4096
#sample_packing: true

adapter: lora
lora_r: 32
lora_alpha: 32
lora_dropout: 0.0  # dropout not supported when using LoRA over expert parameters
lora_target_linear: true

# TODO: not supported for now, see peft#2710xw
#lora_target_parameters:  # target the experts in the last two layers
#  - "22._checkpoint_wrapped_module.mlp.experts.gate_up_proj"
#  - "22._checkpoint_wrapped_module.mlp.experts.down_proj"
#  - "23._checkpoint_wrapped_module.mlp.experts.gate_up_proj"
#  - "23._checkpoint_wrapped_module.mlp.experts.down_proj"

wandb_project: low-stakes-control-sft
wandb_entity: mats-low-stakes
wandb_name: gpt-oss-20b-olympiads-qwen0point6b-malign-prompt-benign-answer-100
wandb_log_model: checkpoint
hub_model_id: EmilRyd/gpt-oss-20b-olympiads-qwen0point6b-malign-prompt-benign-answer-100

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 50 # x4 gpus = 16 batch size

optimizer: adamw_torch_8bit
lr_scheduler: constant_with_warmup
learning_rate: 3e-5

bf16: true
tf32: true

flash_attention: true
attn_implementation: kernels-community/vllm-flash-attn3

gradient_checkpointing: true
activation_offloading: true

logging_steps: 1
save_strategy: best
metric_for_best_model: eval_loss
save_only_model: true
warmup_ratio: 0.01
eval_steps: 10

special_tokens:
eot_tokens:
  - "<|end|>"
```

</details><br>

[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/mats-low-stakes/low-stakes-control-sft/runs/npb88rn7)
# gpt-oss-20b-olympiads-qwen0point6b-malign-prompt-benign-answer-100

This model is a fine-tuned version of [openai/gpt-oss-20b](https://huggingface.co/openai/gpt-oss-20b) on an unknown dataset.

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- total_train_batch_size: 16
- total_eval_batch_size: 16
- optimizer: Use adamw_torch_8bit with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: constant_with_warmup
- lr_scheduler_warmup_steps: 3
- training_steps: 313

### Framework versions

- PEFT 0.17.0
- Transformers 4.55.2
- Pytorch 2.6.0+cu124
- Datasets 4.0.0
- Tokenizers 0.21.4