README.md

---
library_name: transformers
license: gemma
base_model: google/gemma-3-27b-it
tags:
- generated_from_trainer
datasets:
- shisa-ai/shisa-v2-best-of-n-athenev2-tulu70b-llama33-only-no-sysprompt
- shisa-ai/shisa-v2-roleplaying-sft
- shisa-ai/translation_set_april_6
- shisa-ai/rewild-set-deepseek-subset
- shisa-ai/magpie-ultra-set
- shisa-ai/magpie-advanced-questions-set
- shisa-ai/japan-magpie-set
- shisa-ai/shisa-v2-instruction-following-sft
model-index:
- name: outputs/ablation-196-finalsft2-shisa-v2-gemma3-27b
  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: `0.8.0.dev0`
```yaml
# train w/ shisa-ai/shisa-v1-athenev2-reannotated-filtered

base_model: google/gemma-3-27b-it

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: shisa-ai/shisa-v2-best-of-n-athenev2-tulu70b-llama33-only-no-sysprompt
    type: chat_template
    field_messages: conversations
    message_field_role: from
    message_field_content: value
  - path: shisa-ai/shisa-v2-roleplaying-sft
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/translation_set_april_6
    split: train[:25%]
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/rewild-set-deepseek-subset
    split: train[:25%]
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/magpie-ultra-set
    split: train[:8%]
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/magpie-advanced-questions-set
    split: train[:8%]
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/japan-magpie-set
    split: train
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
  - path: shisa-ai/shisa-v2-instruction-following-sft
    split: train[:50%]
    type: chat_template
    field_messages: conversations
    message_property_mappings:
      role: role
      content: content
    roles:
      system:
        - system
      assistant:
        - gpt
        - model
        - assistant
      user:
        - human
        - user
    roles_to_train: ["assistant"]
    
dataset_prepared_path: last_run_prepared
val_set_size: 0.05
output_dir: ./outputs/ablation-196-finalsft2-shisa-v2-gemma3-27b

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

# marginal difference
neftune_noise_alpha: 5

use_wandb: true
wandb_project: shisa-v2
wandb_entity: augmxnt
wandb_name: ablation-196-finalsft2-shisa-v2-gemma3-27b

gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: linear
learning_rate: 5.4e-6

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
evals_per_epoch: 2
eval_table_size:
saves_per_epoch: 0
save_total_limit: 1 # Only store a single checkpoint
debug:
deepspeed: zero3_bf16.json
weight_decay: 1e-4
fsdp:
fsdp_config:
special_tokens:

```

</details><br>

# outputs/ablation-196-finalsft2-shisa-v2-gemma3-27b

This model is a fine-tuned version of [google/gemma-3-27b-it](https://huggingface.co/google/gemma-3-27b-it) on the shisa-ai/shisa-v2-best-of-n-athenev2-tulu70b-llama33-only-no-sysprompt, the shisa-ai/shisa-v2-roleplaying-sft, the shisa-ai/translation_set_april_6, the shisa-ai/rewild-set-deepseek-subset, the shisa-ai/magpie-ultra-set, the shisa-ai/magpie-advanced-questions-set, the shisa-ai/japan-magpie-set and the shisa-ai/shisa-v2-instruction-following-sft datasets.
It achieves the following results on the evaluation set:
- Loss: 0.5417

## 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: 5.4e-06
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- distributed_type: multi-GPU
- num_devices: 64
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 3.0

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 11.3287       | 0.0027 | 1    | 5.4644          |
| 1.1409        | 0.4993 | 182  | 0.5644          |
| 1.0588        | 0.9986 | 364  | 0.5344          |
| 0.9207        | 1.4966 | 546  | 0.5322          |
| 0.8979        | 1.9959 | 728  | 0.5245          |
| 0.7673        | 2.4938 | 910  | 0.5432          |
| 0.7521        | 2.9931 | 1092 | 0.5417          |


### Framework versions

- Transformers 4.50.0
- Pytorch 2.6.0+cu124
- Datasets 3.4.1
- Tokenizers 0.21.1