---
library_name: peft
license: gemma
base_model: huihui-ai/Huihui-gemma-3n-E4B-it-abliterated
tags:
- axolotl
- base_model:adapter:huihui-ai/Huihui-gemma-3n-E4B-it-abliterated
- lora
- transformers
datasets:
- hardlyworking/HardlyRPv2-10k
pipeline_tag: text-generation
model-index:
- name: outputs/out
  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: `0.12.0.dev0`
```yaml
base_model: huihui-ai/Huihui-gemma-3n-E4B-it-abliterated

# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

plugins:
  - axolotl.integrations.liger.LigerPlugin
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
liger_rope: true
liger_rms_norm: true
liger_layer_norm: true
liger_glu_activation: true
liger_fused_linear_cross_entropy: false
cut_cross_entropy: true

load_in_8bit: false
load_in_4bit: true

# for use with fft to only train on language model layers
# unfrozen_parameters:
  # - model.language_model.*
  # - lm_head
  # - embed_tokens


chat_template: gemma3n
eot_tokens:
  - <end_of_turn>
datasets:
  - path: hardlyworking/HardlyRPv2-10k
    type: chat_template
    split: train
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

val_set_size: 0.0
output_dir: ./outputs/out

adapter: qlora
lora_r: 128
lora_alpha: 64
lora_dropout: 0.05
# lora_target_linear: # Does not work with gemma3n currently
lora_target_modules:
  - self_attn.q_proj
  - self_attn.k_proj
  - self_attn.v_proj
  - self_attn.o_proj
  - mlp.gate_proj
  - mlp.up_proj
  - mlp.down_proj


sequence_len: 8192
sample_packing: true
eval_sample_packing: true
pad_to_sequence_len: true

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 4
num_epochs: 2
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

bf16: auto
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
  unsloth: true
resume_from_checkpoint:
logging_steps: 1
# flash_attention: true  # Any attention impl does not work with gemma3n now

warmup_ratio: 0.1
evals_per_epoch:
saves_per_epoch: 1
weight_decay: 0.0
special_tokens:

```

</details><br>

# outputs/out

This model is a fine-tuned version of [huihui-ai/Huihui-gemma-3n-E4B-it-abliterated](https://huggingface.co/huihui-ai/Huihui-gemma-3n-E4B-it-abliterated) on the hardlyworking/HardlyRPv2-10k 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: 0.0002
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 13
- training_steps: 132

### Training results



### Framework versions

- PEFT 0.17.0
- Transformers 4.55.0
- Pytorch 2.7.1+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4