---
library_name: transformers
license: apache-2.0
base_model: timarni/qwen3_pretrain_wiki
tags:
- generated_from_trainer
datasets:
- timarni/MNLP_dataset_mmlu_train
- timarni/sciq_alpaca
model-index:
- name: outputs/qwen3_wiki_sciq_mmlu
  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: `0.9.2`
```yaml
base_model: timarni/qwen3_pretrain_wiki
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
strict: false

chat_template: qwen3
datasets:
  - path: timarni/MNLP_dataset_mmlu_train
    type: alpaca
    split: train
  - path: timarni/sciq_alpaca
    type: alpaca
    split: train

val_set_size: 0.1
output_dir: ./outputs/qwen3_wiki_sciq_mmlu
dataset_prepared_path: last_run_prepared

sequence_len: 4096 #2048
sample_packing: true # was true -> need to check if it actually learns on the samples or not (better understand te hyperparam and event. install axolotl to debug)
eval_sample_packing: true
pad_to_sequence_len: true

# To be sure that no LORA is done
adapter: null
lora: false
merge_lora: false

wandb_project: mnlp_project
wandb_entity: tim-arni
wandb_watch:
wandb_name: qwen3_wiki_sciq_mmlu
wandb_log_model:

gradient_accumulation_steps: 16 # 2
micro_batch_size: 2 # 1
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00005 # 0.00005

bf16: auto
tf32: true

gradient_checkpointing: offload
gradient_checkpointing_kwargs:
  use_reentrant: false
resume_from_checkpoint:
logging_steps: 1
gradient_clipping: 1.0
flash_attention: true

warmup_steps: 20
evals_per_epoch: 4
saves_per_epoch: 1
weight_decay: 0.01
special_tokens:

```

</details><br>

# outputs/qwen3_wiki_sciq_mmlu

This model is a fine-tuned version of [timarni/qwen3_pretrain_wiki](https://huggingface.co/timarni/qwen3_pretrain_wiki) on the timarni/MNLP_dataset_mmlu_train and the timarni/sciq_alpaca datasets.
It achieves the following results on the evaluation set:
- Loss: 0.0664

## 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-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- total_eval_batch_size: 8
- optimizer: Use 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: 20
- num_epochs: 3.0

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 0.3628        | 0.0147 | 1    | 0.3654          |
| 0.0724        | 0.25   | 17   | 0.0709          |
| 0.0601        | 0.5    | 34   | 0.0625          |
| 0.0586        | 0.75   | 51   | 0.0589          |
| 0.055         | 1.0    | 68   | 0.0563          |
| 0.0344        | 1.25   | 85   | 0.0580          |
| 0.0279        | 1.5    | 102  | 0.0614          |
| 0.0287        | 1.75   | 119  | 0.0615          |
| 0.0333        | 2.0    | 136  | 0.0598          |
| 0.0188        | 2.25   | 153  | 0.0617          |
| 0.0156        | 2.5    | 170  | 0.0659          |
| 0.0176        | 2.75   | 187  | 0.0663          |
| 0.0277        | 3.0    | 204  | 0.0664          |


### Framework versions

- Transformers 4.51.3
- Pytorch 2.5.1+cu121
- Datasets 3.5.1
- Tokenizers 0.21.1