See axolotl config

axolotl version: 0.6.0

base_model: Qwen/Qwen2-7B
trust_remote_code: true

strict: false

chat_template: llama3
datasets:
  - path: penfever/allenai_WildChat-1M-Full-meta-llama_Llama-3.1-8B-Instruct
    type: chat_template
    split: train[:25%]
    field_messages: conversation
    message_field_role: role
    message_field_content: content

dataset_prepared_path: /scratch/bf996/axolotl/datasets/wildchat-250k-llama-3.1-8b-instruct
val_set_size: 0.02
output_dir: /scratch/bf996/axolotl/outputs/qwen-2-7b-wildchat-250k-llama-3.1-8b-instruct

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

wandb_project: lm-evals
wandb_entity:
wandb_watch:
wandb_name: qwen-2-7b-WildChat-llama-3.1-8b-instruct
wandb_log_model:
hub_model_id: penfever/qwen-2-7b-WildChat-250k-llama-3.1-8b-instruct


gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 2e-5

train_on_inputs: false
group_by_length: false
bf16: auto
fp16:
tf32: false

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
evals_per_epoch: 0
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  pad_token: <|finetune_right_pad_id|>
  eos_token: <|eot_id|>
  bos_token: <|begin_of_text|>

qwen-2-7b-WildChat-250k-llama-3.1-8b-instruct

This model is a fine-tuned version of Qwen/Qwen2-7B on the penfever/allenai_WildChat-1M-Full-meta-llama_Llama-3.1-8B-Instruct dataset. It achieves the following results on the evaluation set:

• Loss: 6.7290

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: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• total_eval_batch_size: 4
• optimizer: Use paged_adamw_8bit 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: 100
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
8.1514	0.9998	2815	6.7290

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0