You can deploy this model and start earning money today!

wangd12/
financebench_llama_3_1_8b_8bits_r8_dora

Name: financebench_llama_3_1_8b_8bits_r8_dora
Availability: InStock

safetensors

Views

No views yet

See axolotl config

axolotl version: 0.9.1

yaml

base_model: meta-llama/Llama-3.1-8B-Instruct
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer
gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0001
load_in_8bit: true
load_in_4bit: false
adapter: lora
lora_model_dir: null
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
lora_target_modules:
- q_proj
- v_proj
- k_proj
datasets:
- path: /workspace/FinLoRA/data/train/financebench_train.jsonl
  type:
    system_prompt: ''
    field_system: system
    field_instruction: context
    field_output: target
    format: '[INST] {instruction} [/INST]'
    no_input_format: '[INST] {instruction} [/INST]'
dataset_prepared_path: null
val_set_size: 0.02
output_dir: /workspace/FinLoRA/lora/axolotl-output/financebench_llama_3_1_8b_8bits_r8_dora
peft_use_dora: true
peft_use_rslora: false
sequence_len: 4096
sample_packing: false
pad_to_sequence_len: false
wandb_project: finlora_models
wandb_entity: null
wandb_watch: gradients
wandb_name: financebench_llama_3_1_8b_8bits_r8_dora
wandb_log_model: 'false'
bf16: auto
tf32: false
gradient_checkpointing: true
resume_from_checkpoint: null
logging_steps: 500
flash_attention: false
deepspeed: deepspeed_configs/zero1.json
warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1
weight_decay: 0.0
special_tokens:
  pad_token: <|end_of_text|>
chat_template: llama3

workspace/FinLoRA/lora/axolotl-output/financebench_llama_3_1_8b_8bits_r8_dora

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the /workspace/FinLoRA/data/train/financebench_train.jsonl dataset. It achieves the following results on the evaluation set:

Loss: 3.0309

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: 0.0001
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 5
gradient_accumulation_steps: 2
total_train_batch_size: 10
total_eval_batch_size: 5
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: 10
num_epochs: 4.0

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.1176	1	4.7299
No log	0.2353	2	4.7794
No log	0.4706	4	4.7367
No log	0.7059	6	4.6608
No log	0.9412	8	4.5025
No log	1.1176	10	4.2269
No log	1.3529	12	3.9951
No log	1.5882	14	3.7253
No log	1.8235	16	3.5911
No log	2.0	18	3.4499
No log	2.2353	20	3.4191
No log	2.4706	22	3.2604
No log	2.7059	24	3.1056
No log	2.9412	26	2.9699
No log	3.1176	28	3.0718
No log	3.3529	30	3.0529
No log	3.5882	32	3.0309

Framework versions

PEFT 0.15.2
Transformers 4.51.3
Pytorch 2.8.0.dev20250319+cu128
Datasets 3.5.1
Tokenizers 0.21.1

wangd12/financebench_llama_3_1_8b_8bits_r8_dora

workspace/FinLoRA/lora/axolotl-output/financebench_llama_3_1_8b_8bits_r8_dora

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

wangd12/
financebench_llama_3_1_8b_8bits_r8_dora