---
library_name: peft
license: llama3.1
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
tags:
- axolotl
- generated_from_trainer
model-index:
- name: harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b
  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/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: `0.4.1`
```yaml
# Configure the base model 

strict: false
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
tokenizer_config: meta-llama/Meta-Llama-3.1-8B-Instruct
model_type: AutoModelForCausalLM


# Output configuration
hub_model_id: collinear-ai/harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b
dataset_prepared_path: /workspace/gen_judge/data/collinear-ai/harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b
output_dir: /workspace/gen_judge/collinear-ai/harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b

# Format the dataset into the right instruction format.
chat_template: llama3  #llama 3 instruct chat template USE
datasets:
  - path: collinear-ai/prompt-response-eval-classification-dataset-final-axolotl
    split: without_animal_env_abuse_train
    type: chat_template
    chat_template: llama3
    field_messages: train_conv
    message_field_role: role
    message_field_content: content
train_on_inputs: false #FALSE

val_set_size: 0.05
# Data packing
sequence_len: 2048
eval_sample_packing: false
sample_packing: false
pad_to_sequence_len: true
group_by_length: false

# Lora config
adapter: qlora
lora_model_dir:
load_in_8bit: false
load_in_4bit: true
lora_r: 64
lora_alpha: 32
lora_dropout: 0.1
lora_target_linear: true
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj
lora_modules_to_save: 
  - embed_tokens
  - lm_head

# Logging config
wandb_project: general-judge-harmfulness-bif-data
wandb_entity: nazneen
wandb_name: collinear-ai/harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b

# Trainer config
gradient_accumulation_steps: 2
micro_batch_size: 12
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 0.00005

bfloat16: true
bf16: true
fp16:
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 10
xformers_attention:
flash_attention: true
save_safetensors: true

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 50
evals_per_epoch: 3
eval_table_size:
eval_max_new_tokens: 500
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.02
fsdp_config:
special_tokens:
  pad_token: "<|end_of_text|>"


## weight decay
## add validation set (split add)

```

</details><br>

# harm-category-labelled-multi-task-16cat-26_12-leaveone-without-animal_env_abuse-llama8b

This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 0.1492

## 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: 12
- eval_batch_size: 12
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 2
- total_train_batch_size: 48
- total_eval_batch_size: 24
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 50
- num_epochs: 3

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| No log        | 0.0007 | 1    | 1.8779          |
| 0.0152        | 0.3332 | 461  | 0.1341          |
| 0.0072        | 0.6664 | 922  | 0.1252          |
| 0.0046        | 0.9996 | 1383 | 0.1301          |
| 0.0022        | 1.3329 | 1844 | 0.1315          |
| 0.0028        | 1.6661 | 2305 | 0.1361          |
| 0.0018        | 1.9993 | 2766 | 0.1338          |
| 0.0005        | 2.3325 | 3227 | 0.1465          |
| 0.0003        | 2.6657 | 3688 | 0.1488          |
| 0.0002        | 2.9989 | 4149 | 0.1492          |


### Framework versions

- PEFT 0.11.1
- Transformers 4.45.0
- Pytorch 2.1.2+cu118
- Datasets 2.19.1
- Tokenizers 0.20.3