---
library_name: transformers
license: gemma
base_model: google/gemma-3-270m-it
tags:
- generated_from_trainer
datasets:
- HuggingFaceH4/CodeAlpaca_20K
model-index:
- name: outputs/gemma-3-270m-codealpaca-finetune
  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: google/gemma-3-270m-it
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

ddp_find_unused_parameters: true

load_in_8bit: false
load_in_4bit: false

chat_template: gemma3
eot_tokens:
  - "<end_of_turn>"

datasets:
  - path: HuggingFaceH4/CodeAlpaca_20K
    type:
      field_instruction: prompt
      field_input: input
      field_output: output
      format: |
        Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.

        ### Instruction:
        {instruction}

        ### Input:
        {input}

        ### Response:
      no_input_format: |
        Below is an instruction that describes a task. Write a response that appropriately completes the request.

        ### Instruction:
        {instruction}

        ### Response:

val_set_size: 0.05 # Use 5% of the data for validation
output_dir: ./outputs/gemma-3-270m-codealpaca-finetune

sequence_len: 2048
sample_packing: true
eval_sample_packing: false

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00002

bf16: true
tf32: true

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false

resume_from_checkpoint:
logging_steps: 1
flash_attention: true

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

```

</details><br>

# outputs/gemma-3-270m-codealpaca-finetune

This model is a fine-tuned version of [google/gemma-3-270m-it](https://huggingface.co/google/gemma-3-270m-it) on the HuggingFaceH4/CodeAlpaca_20K dataset.
It achieves the following results on the evaluation set:
- Loss: nan
- Memory/max Memory Active(gib): 8.51
- Memory/max Memory Allocated(gib): 8.51
- Memory/device Memory Reserved(gib): 10.27

## 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: 2e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Use OptimizerNames.ADAMW_TORCH 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: 34
- training_steps: 348

### Training results

| Training Loss | Epoch  | Step | Validation Loss | Memory Active(gib) | Memory Allocated(gib) | Memory Reserved(gib) |
|:-------------:|:------:|:----:|:---------------:|:------------------:|:---------------------:|:--------------------:|
| No log        | 0      | 0    | nan             | 5.84               | 5.84                  | 5.86                 |
| 0.0           | 0.9978 | 116  | nan             | 8.51               | 8.51                  | 10.27                |
| 0.0           | 1.9892 | 232  | nan             | 8.51               | 8.51                  | 10.27                |
| 0.0           | 2.9806 | 348  | nan             | 8.51               | 8.51                  | 10.27                |


### Framework versions

- Transformers 4.55.0
- Pytorch 2.6.0+cu126
- Datasets 4.0.0
- Tokenizers 0.21.4