metadata

base_model: Qwen/Qwen2.5-14B-Instruct
library_name: peft
license: apache-2.0
tags:
  - generated_from_trainer
model-index:
  - name: outputs/lora-out
    results: []

See axolotl config

axolotl version: 0.4.1

base_model: Qwen/Qwen2.5-14B-Instruct

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: output.jsonl
    type:
      field_instruction: instruction
      field_input: input
      field_output: output
      format: "<|im_start|>system\n{instruction}<|im_end|>\n<|im_start|>user\n{input}<|im_end|>\n<|im_start|>assistant\n"

special_tokens:
  bos_token:
  eos_token: "<|im_end|>"
  pad_token: "<|endoftext|>"

dataset_prepared_path:
val_set_size: 0.05
output_dir: ./outputs/lora-out

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 8
lora_alpha: 16
lora_dropout: 0.05
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

wandb_project: mssong_axolotl
wandb_entity: mssong
wandb_watch:
wandb_run_id: 
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 3
optimizer:
lr_scheduler: cosine
learning_rate: 0.00005
train_on_inputs:
group_by_length: false
bf16: true
fp16: false
tf32: false
gradient_checkpointing: true
early_stopping_patience: 3
local_rank:
logging_steps: 10
xformers_attention:
flash_attention: true
#warmup_ratio: 0.02
warmup_steps: 100
eval_steps: 100
save_steps: 500
save_total_limit: 2
eval_sample_packing:
debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
trust_remote_code: true

outputs/lora-out

This model is a fine-tuned version of Qwen/Qwen2.5-14B-Instruct on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.0405

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: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 2
total_train_batch_size: 4
total_eval_batch_size: 2
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0035	1	0.6979
0.046	0.3515	100	0.0793
0.0259	0.7030	200	0.0519
0.0242	1.0545	300	0.0447
0.0194	1.4060	400	0.0435
0.016	1.7575	500	0.0427
0.0097	2.1090	600	0.0392
0.0179	2.4605	700	0.0410
0.0081	2.8120	800	0.0405

Framework versions

PEFT 0.13.0
Transformers 4.45.1
Pytorch 2.3.1+cu121
Datasets 2.21.0
Tokenizers 0.20.0