See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen1.5-14B-Chat
bf16: auto
dataset_prepared_path: null
datasets:
- data_files:
  - f70ddae1849231d5_train_data.json
  ds_type: json
  format: custom
  path: f70ddae1849231d5_train_data.json
  type:
    field: null
    field_input: null
    field_instruction: instruction
    field_output: response
    field_system: null
    format: null
    no_input_format: null
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
evals_per_epoch: 2
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
group_by_length: false
hub_model_id: taopanda-1/a1dedc9d-b040-4f80-9e8b-ef01e5dc25a1
learning_rate: 0.0002
load_in_4bit: false
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: true
lr_scheduler: cosine
micro_batch_size: 1
num_epochs: 1
optimizer: paged_adamw_8bit
output_dir: ./outputs/out/taopanda-1_626c7685-3b25-4cd2-a8a4-a8e58ec0f209
pad_to_sequence_len: false
resume_from_checkpoint: null
sample_packing: false
saves_per_epoch: 1
seed: 63259
sequence_len: 1024
special_tokens: null
strict: false
tf32: true
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: fatcat87-taopanda
wandb_log_model: null
wandb_mode: online
wandb_name: taopanda-1_626c7685-3b25-4cd2-a8a4-a8e58ec0f209
wandb_project: subnet56
wandb_runid: taopanda-1_626c7685-3b25-4cd2-a8a4-a8e58ec0f209
wandb_watch: null
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

a1dedc9d-b040-4f80-9e8b-ef01e5dc25a1

This model is a fine-tuned version of Qwen/Qwen1.5-14B-Chat on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2882

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 1
• seed: 63259
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• total_eval_batch_size: 4
• 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
0.6166	0.0004	1	0.5929
0.2901	0.4999	1419	0.2989
0.34	0.9998	2838	0.2882

Framework versions

• PEFT 0.11.1
• Transformers 4.42.3
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1