OsakanaTeishoku/Qwen2.5-7B-axolotl-sft-v0.1

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-7B",
    quantization_config=BitsAndBytesConfig(
        load_in_4bit=True,
    ),
)
model = PeftModel.from_pretrained(base_model, "OsakanaTeishoku/Qwen2.5-7B-axolotl-sft-v0.1")
tokenizer = AutoTokenizer.from_pretrained("OsakanaTeishoku/Qwen2.5-7B-axolotl-sft-v0.1")

from transformers import TextStreamer

streamer = TextStreamer(
    tokenizer,
    skip_prompt=False, 
    skip_special_tokens=False, 
)

prompt = "あなたは何者ですか"
messages = [
    {"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512,
    streamer=streamer,
    eos_token_id=tokenizer.eos_token_id,
    pad_token_id=tokenizer.eos_token_id,
)

base_model: Qwen/Qwen2.5-7B
hub_model_id: OsakanaTeishoku/custom_model_name

load_in_8bit: false
load_in_4bit: true
strict: false

chat_template: qwen_25

datasets:
  # This will be the path used for the data when it is saved to the Volume in the cloud.
  - path: Aratako/Magpie-Tanuki-8B-annotated-96k
    split: train[0:4000]
    type: chat_template
    field_messages: messages


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

sequence_len: 2048
sample_packing: false
eval_sample_packing: false
pad_to_sequence_len: false

adapter: qlora
lora_model_dir:
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
lora_modules_to_save: # required when adding new tokens to LLaMA/Mistral
  - embed_tokens
  - lm_head

gradient_accumulation_steps: 1
micro_batch_size: 16
num_epochs: 1
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.0001

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 10
save_steps:
debug:
deepspeed: /workspace/axolotl/deepspeed_configs/zero2.json
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|im_end|>"

plugins:
  - axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_layer_norm: true
liger_fused_linear_cross_entropy: true

eval_strategy: "no"
save_strategy: "epoch"