There is no model saved in the checkpoint/final when lora_tuning llama3

[leoozy]

Please check that this issue hasn't been reported before.

• [X] I searched previous Bug Reports didn't find any similar reports.

Expected Behavior

Save pytorch_model.bin at the output file

Current behaviour

There is only model_adapter.bin. No model saved

Steps to reproduce

command = f"/opt/anaconda3/envs/open-sora/bin/accelerate launch -m axolotl.cli.train {args.yaml_path} --deepspeed {args.deepspeed}"

Config yaml

base_model: /zhangjunlei/download/models/Meta-Llama-3-8B-Instruct
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: /zhangjunlei/download/data/webagent/axo_vwa_cot_json_v3.json
    ds_type: json
    type: alpaca
dataset_prepared_path:
val_set_size: 0.05
output_dir: /zhangjunlei/code/axolotl/output/llama3.1_70b_lora_vwa_0801_1_debug3

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 512
lora_alpha: 1024
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 2
micro_batch_size: 1
num_epochs: 4
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
s2_attention:

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 8
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  eos_token: "<|eot_id|>"
  pad_token: "<|end_of_text|>"
eval_sample_packing: False

Possible solution

No response

Which Operating Systems are you using?

• [X] Linux
• [ ] macOS
• [ ] Windows

Python Version

3.10

axolotl branch-commit

main

Acknowledgements

• [X] My issue title is concise, descriptive, and in title casing.
• [X] I have searched the existing issues to make sure this bug has not been reported yet.
• [X] I am using the latest version of axolotl.
• [X] I have provided enough information for the maintainers to reproduce and diagnose the issue.