---
license: apache-2.0
base_model: HuggingFaceTB/cosmo-1b
tags:
- generated_from_trainer
model-index:
- name: galore-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. -->

Catastrophic forgetting test results:

Initial evaluation loss on 1k subset of HuggingFaceTB/cosmopedia-100k dataset was 1.182. 100 steps of LISA training reduced this to 1.117.

Comparison to control: cosmo-1b started out with 1.003 loss on (a different subset of) dataset, increasing to 1.024 at 100 steps.

[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.4.0`
```yaml
base_model: HuggingFaceTB/cosmo-1b
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: Vezora/Tested-22k-Python-Alpaca
    type: alpaca
dataset_prepared_path: prepared-qlora
val_set_size: 0.05
output_dir: ./galore-out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

adapter:
lora_model_dir:
lora_r:
lora_alpha:
lora_dropout:
lora_target_linear:
lora_fan_in_fan_out:

wandb_project: cosmo-python-galore
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 1
optimizer: galore_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.0005

optim_target_modules:
  - self_attn  # for llama
  - mlp

optim_args:
  rank: 256
  scale: 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: 10
evals_per_epoch: 4
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

```

</details><br>

# galore-out

This model is a fine-tuned version of [HuggingFaceTB/cosmo-1b](https://huggingface.co/HuggingFaceTB/cosmo-1b) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.2426

## 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: 0.0005
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 1

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 0.6299        | 0.0   | 1    | 0.6469          |
| 0.4353        | 0.25  | 215  | 0.4139          |
| 0.3721        | 0.5   | 430  | 0.2957          |
| 0.3514        | 0.75  | 645  | 0.2426          |


### Framework versions

- Transformers 4.40.0.dev0
- Pytorch 2.1.2+cu118
- Datasets 2.18.0
- Tokenizers 0.15.0