---
library_name: transformers
base_model: NewEden/Hamanasu-KTO-4B
tags:
- axolotl
- generated_from_trainer
datasets:
- PocketDoc/Dans-Prosemaxx-Cowriter-3-S
- PocketDoc/Dans-Prosemaxx-Adventure
- PocketDoc/Dans-Failuremaxx-Adventure-3
- PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-2
- PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-3
- PocketDoc/Dans-Prosemaxx-InstructWriter-Continue-2
- PocketDoc/Dans-Prosemaxx-Instructwriter-Long
model-index:
- name: Hamanasu-4B-Adventure-Final-Hopefully
  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
base_model: NewEden/Hamanasu-KTO-4B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

hub_model_id: NewEden/Hamanasu-4B-Adventure-Final-Hopefully
hub_strategy: "all_checkpoints"
push_dataset_to_hub:
hf_use_auth_token: true

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: false
strict: false

datasets:
  - path: PocketDoc/Dans-Prosemaxx-Cowriter-3-S
    type: dan-chat-advanced
  - path: PocketDoc/Dans-Prosemaxx-Adventure
    type: dan-chat-advanced
  - path: PocketDoc/Dans-Failuremaxx-Adventure-3
    type: dan-chat-advanced
  - path: PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-2
    type: dan-chat-advanced  
  - path: PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-3
    type: dan-chat-advanced  
  - path: PocketDoc/Dans-Prosemaxx-InstructWriter-Continue-2
    type: dan-chat-advanced  
  - path: PocketDoc/Dans-Prosemaxx-Instructwriter-Long
    type: dan-chat-advanced  
shuffle_merged_datasets: true
dataset_prepared_path: prepared_data
val_set_size: 0.01
output_dir: ./adventure

sequence_len: 32768
sample_packing: true
pad_to_sequence_len: true
eval_sample_packing: true


wandb_project: tavbussy
wandb_entity:
wandb_watch:
wandb_name: adventure-attempt-02
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 5e-6
max_grad_norm: 0.1

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 25
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 2
debug:
deepspeed: ./deepspeed_configs/zero3_bf16.json
weight_decay: 0.02
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|finetune_right_pad_id|>

```

</details><br>

# Hamanasu-4B-Adventure-Final-Hopefully

This model is a fine-tuned version of [NewEden/Hamanasu-KTO-4B](https://huggingface.co/NewEden/Hamanasu-KTO-4B) on the PocketDoc/Dans-Prosemaxx-Cowriter-3-S, the PocketDoc/Dans-Prosemaxx-Adventure, the PocketDoc/Dans-Failuremaxx-Adventure-3, the PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-2, the PocketDoc/Dans-Prosemaxx-InstructWriter-ZeroShot-3, the PocketDoc/Dans-Prosemaxx-InstructWriter-Continue-2 and the PocketDoc/Dans-Prosemaxx-Instructwriter-Long datasets.
It achieves the following results on the evaluation set:
- Loss: 2.4143

## 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: 5e-06
- 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 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: 25
- num_epochs: 4.0

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 2.5668        | 0.0068 | 1    | 2.5806          |
| 2.455         | 0.2534 | 37   | 2.4450          |
| 2.4115        | 0.5068 | 74   | 2.4323          |
| 2.3298        | 0.7603 | 111  | 2.4223          |
| 2.322         | 1.0137 | 148  | 2.4178          |
| 2.2661        | 1.2671 | 185  | 2.4178          |
| 2.2482        | 1.5205 | 222  | 2.4155          |
| 2.3707        | 1.7740 | 259  | 2.4115          |
| 2.293         | 2.0274 | 296  | 2.4132          |
| 2.3085        | 2.2808 | 333  | 2.4137          |
| 2.1902        | 2.5342 | 370  | 2.4123          |
| 2.216         | 2.7877 | 407  | 2.4112          |
| 2.3081        | 3.0411 | 444  | 2.4123          |
| 2.1989        | 3.2945 | 481  | 2.4142          |
| 2.2527        | 3.5479 | 518  | 2.4142          |
| 2.2419        | 3.8014 | 555  | 2.4143          |


### Framework versions

- Transformers 4.50.0
- Pytorch 2.5.1+cu124
- Datasets 3.4.1
- Tokenizers 0.21.1