"KeyError: 'conversation'" when loading config

[T145]

This config:

base_model: T145/ZEUS-8B-V2
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: true
load_in_4bit: false
strict: false

chat_template: llama3
rl: dpo
datasets:
  - path: fozziethebeat/alpaca_messages_2k_dpo_test
    type: chat_template.default
    field_messages: conversation
    field_chosen: chosen
    field_rejected: rejected
    message_field_role: role
    message_field_content: content
    roles:
      system:
        - system
      user:
        - user
      assistant:
        - assistant

dataset_prepared_path:
val_set_size: 0
output_dir: ./outputs/lora-out

sequence_len: 4096
sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

gradient_accumulation_steps: 4
micro_batch_size: 2
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: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:

Creates this error:

[2024-12-14 08:38:12,907] [DEBUG] [axolotl.load_tokenizer:296] [PID:329] [RANK:0] EOS: 128009 / <|eot_id|>
[2024-12-14 08:38:12,908] [DEBUG] [axolotl.load_tokenizer:297] [PID:329] [RANK:0] BOS: 128000 / <|begin_of_text|>  
[2024-12-14 08:38:12,908] [DEBUG] [axolotl.load_tokenizer:298] [PID:329] [RANK:0] PAD: 128004 / <|finetune_right_pad_id|>
[2024-12-14 08:38:12,908] [DEBUG] [axolotl.load_tokenizer:299] [PID:329] [RANK:0] UNK: None / None
Mapping RL Dataset:   0%|                                                         | 0/44245 [00:00<?, ? examples/s]Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/workspace/axolotl/src/axolotl/cli/train.py", line 58, in <module>
    fire.Fire(do_cli)
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 135, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 468, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
                                ^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/fire/core.py", line 684, in _CallAndUpdateTrace  
    component = fn(*varargs, **kwargs)
                ^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/cli/train.py", line 34, in do_cli
    return do_train(parsed_cfg, parsed_cli_args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/cli/train.py", line 43, in do_train
    dataset_meta = load_rl_datasets(cfg=cfg, cli_args=cli_args)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/cli/__init__.py", line 515, in load_rl_datasets
    train_dataset, eval_dataset = load_prepare_dpo_datasets(cfg)
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/utils/data/rl.py", line 195, in load_prepare_dpo_datasets
    train_dataset = load_split(cfg.datasets, cfg)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/utils/data/rl.py", line 153, in load_split
    split_datasets[i] = map_dataset(
                        ^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/utils/data/rl.py", line 70, in map_dataset
    data_set = data_set.map(
               ^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 560, in wrapper 
    out: Union["Dataset", "DatasetDict"] = func(self, *args, **kwargs)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 3055, in map    
    for rank, done, content in Dataset._map_single(**dataset_kwargs):
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 3428, in _map_single
    example = apply_function_on_filtered_inputs(example, i, offset=offset)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/datasets/arrow_dataset.py", line 3320, in apply_function_on_filtered_inputs
    processed_inputs = function(*fn_args, *additional_args, **fn_kwargs)
                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/prompt_strategies/dpo/chat_template.py", line 40, in transform_fn
    messages = sample[field_messages]
               ~~~~~~^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/datasets/formatting/formatting.py", line 277, in 
__getitem__
    value = self.data[key]
            ~~~~~~~~~^^^^^
KeyError: 'conversation'
Traceback (most recent call last):
  File "/root/miniconda3/envs/py3.11/bin/accelerate", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/accelerate/commands/accelerate_cli.py", line 48, 
in main
    args.func(args)
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/accelerate/commands/launch.py", line 1168, in launch_command
    simple_launcher(args)
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/accelerate/commands/launch.py", line 763, in simple_launcher
    raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['/root/miniconda3/envs/py3.11/bin/python3', '-m', 'axolotl.cli.train', './outputs/instruct-dpo-lora-8b.yml', '--debug-num-examples', '0']' returned non-zero exit status 1.
Traceback (most recent call last):
  File "/root/miniconda3/envs/py3.11/bin/axolotl", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/workspace/axolotl/src/axolotl/cli/main.py", line 229, in main
    cli()
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/click/core.py", line 1157, in __call__
    return self.main(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/click/core.py", line 1078, in main
    rv = self.invoke(ctx)
         ^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/click/core.py", line 1688, in invoke
    return _process_result(sub_ctx.command.invoke(sub_ctx))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/click/core.py", line 1434, in invoke
    return ctx.invoke(self.callback, **ctx.params)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/site-packages/click/core.py", line 783, in invoke
    return __callback(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/workspace/axolotl/src/axolotl/cli/main.py", line 56, in train
    subprocess.run(cmd, check=True)  # nosec B603
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniconda3/envs/py3.11/lib/python3.11/subprocess.py", line 571, in run
    raise CalledProcessError(retcode, process.args,
subprocess.CalledProcessError: Command '['accelerate', 'launch', '-m', 'axolotl.cli.train', './outputs/instruct-dpo-lora-8b.yml', '--debug-num-examples', '0']' returned non-zero exit status 1.

What's the problem? Does the conversation field need to be defined somewhere else?

[ehartford]

conversations

[T145]

Wdym? The dataset column is conversation: am I missing something?

[ehartford]

Actually it's not.

https://huggingface.co/datasets/fozziethebeat/alpaca_messages_2k_test/viewer

It's "messages"

[T145]

I'm using the DPO test dataset: https://huggingface.co/datasets/fozziethebeat/alpaca_messages_2k_dpo_test

I copied the example from here: https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/llama-3/instruct-dpo-lora-8b.yml

And just replaced the target model, which is why I'm curious as to what's wrong.

[ehartford]

Just look at the dataset and look at your config.

The dataset has "messages"
The config has "conversation"

That's the answer to your question.

As to why the example said "conversation" - likely the dataset changed since the example was written.

[NanoCode012]

Hey @T145 , I ran this exact config https://github.com/axolotl-ai-cloud/axolotl/blob/main/examples/llama-3/instruct-dpo-lora-8b.yml
and I was able to get it to work. Can you verify it works for you?

[NanoCode012]

I ran your exact config and didn't have this issue. What commit are you on?

[T145]

Ironically enough I've been granted access to every Meta model family but V3 xd If I change the model to V3.1 it works. This is command I used:

docker run --gpus '"all"' --rm -it axolotlai/axolotl:main-latest

and the version is: axolotl, version 0.6.0