---
library_name: peft
license: llama3.1
base_model: meta-llama/Llama-3.1-70B-Instruct
tags:
- generated_from_trainer
model-index:
- name: outputs/out/qlora-llama3-70b
  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: `0.5.0`
```yaml
adapter: qlora
base_model: meta-llama/Llama-3.1-70B-Instruct
bf16: true
chat_template: tokenizer_default
datasets:
- field_human: user
  field_messages: messages
  field_model: assistant
  field_system: system
  message_field_content: content
  message_field_role: role
  path: s3://tensorkube-datasets-bucket-3fb7f7dc-da71-4990/vaero-data.jsonl
  type: chat_template
debug: false
deepspeed: null
early_stopping_patience: null
eval_table_size: null
evals_per_epoch: 1
flash_attention: true
fp16: false
fsdp:
- full_shard
- auto_wrap
fsdp_config:
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_cpu_ram_efficient_loading: true
  fsdp_limit_all_gathers: true
  fsdp_offload_params: false
  fsdp_sharding_strategy: FULL_SHARD
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_sync_module_states: true
  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
  fsdp_use_orig_params: false
gradient_accumulation_steps: 2
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
group_by_length: false
hf_org_id: samagra-tensorfuse
learning_rate: 7.0e-05
load_in_4bit: true
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: false
lora_target_modules:
- q_proj
- v_proj
- gate_proj
- up_proj
- k_proj
- down_proj
- o_proj
lr_scheduler: linear
micro_batch_size: 4
model_type: LlamaForCausalLM
num_epochs: 1
optimizer: adamw_torch
output_dir: ./outputs/out/qlora-llama3-70b
pad_to_sequence_len: false
peft_use_rslora: true
resume_from_checkpoint: null
sample_packing: false
saves_per_epoch: 1
sequence_len: 4096
special_tokens:
  pad_token: <|finetune_right_pad_id|>
strict: false
tf32: false
tokenizer_add_special_tokens: true
tokenizer_padding_side: right
tokenizer_type: AutoTokenizer
train_on_inputs: false
val_set_size: 0.1
wandb_entity: samagra14-tensorfuse
wandb_mode: online
wandb_name: 2025-04-03-02
wandb_project: green-test-templatev7432-tok
wandb_watch: none
warmup_steps: 5
weight_decay: 0.1
xformers_attention: null

```

</details><br>

# outputs/out/qlora-llama3-70b

This model is a fine-tuned version of [meta-llama/Llama-3.1-70B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-70B-Instruct) on the None dataset.
It achieves the following results on the evaluation set:
- Loss: 1.6805

## 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: 7e-05
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- total_eval_batch_size: 16
- optimizer: Use OptimizerNames.ADAMW_TORCH 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: 5
- num_epochs: 1

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 1.7623        | 0.9333 | 7    | 1.6805          |


### Framework versions

- PEFT 0.13.2
- Transformers 4.46.1
- Pytorch 2.3.1+cu121
- Datasets 3.0.1
- Tokenizers 0.20.3