AttributeError: module 'parlai.tasks.fromfile.agents' has no attribute 'create_agents'

[mohamedae17]

I try to train ParlAI on custom data set but when i run :

!parlai train_model -t fromfile --fromfile-datapath /content/drive/MyDrive/sam.txt -m transformer/generator -mf /content/drive/MyDrive/sammodel -bs 8 -eps 10

i get this error :

2023-05-03 05:30:01.970264: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-05-03 05:30:02.855068: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
05:30:04 | building dictionary first...
05:30:04 | Opt:
05:30:04 |     activation: relu
05:30:04 |     adafactor_eps: '(1e-30, 0.001)'
05:30:04 |     adam_eps: 1e-08
05:30:04 |     add_p1_after_newln: False
05:30:04 |     aggregate_micro: False
05:30:04 |     allow_missing_init_opts: False
05:30:04 |     attention_dropout: 0.0
05:30:04 |     batchsize: 1
05:30:04 |     beam_block_full_context: True
05:30:04 |     beam_block_list_filename: None
05:30:04 |     beam_block_ngram: -1
05:30:04 |     beam_context_block_ngram: -1
05:30:04 |     beam_delay: 30
05:30:04 |     beam_length_penalty: 0.65
05:30:04 |     beam_min_length: 1
05:30:04 |     beam_size: 1
05:30:04 |     betas: '(0.9, 0.999)'
05:30:04 |     bpe_add_prefix_space: None
05:30:04 |     bpe_debug: False
05:30:04 |     bpe_dropout: None
05:30:04 |     bpe_merge: None
05:30:04 |     bpe_vocab: None
05:30:04 |     checkpoint_activations: False
05:30:04 |     compute_tokenized_bleu: False
05:30:04 |     datapath: /usr/local/lib/python3.10/dist-packages/data
05:30:04 |     datatype: train
05:30:04 |     delimiter: '\n'
05:30:04 |     dict_class: parlai.core.dict:DictionaryAgent
05:30:04 |     dict_endtoken: __end__
05:30:04 |     dict_file: /content/drive/MyDrive/sammodel.dict
05:30:04 |     dict_include_test: False
05:30:04 |     dict_include_valid: False
05:30:04 |     dict_initpath: None
05:30:04 |     dict_language: english
05:30:04 |     dict_loaded: False
05:30:04 |     dict_lower: False
05:30:04 |     dict_max_ngram_size: -1
05:30:04 |     dict_maxexs: -1
05:30:04 |     dict_maxtokens: -1
05:30:04 |     dict_minfreq: 0
05:30:04 |     dict_nulltoken: __null__
05:30:04 |     dict_starttoken: __start__
05:30:04 |     dict_textfields: text,labels
05:30:04 |     dict_tokenizer: re
05:30:04 |     dict_unktoken: __unk__
05:30:04 |     display_examples: False
05:30:04 |     download_path: None
05:30:04 |     dropout: 0.0
05:30:04 |     dynamic_batching: None
05:30:04 |     embedding_projection: random
05:30:04 |     embedding_size: 300
05:30:04 |     embedding_type: random
05:30:04 |     embeddings_scale: True
05:30:04 |     eval_batchsize: None
05:30:04 |     eval_dynamic_batching: None
05:30:04 |     evaltask: None
05:30:04 |     ffn_size: 300
05:30:04 |     final_extra_opt: 
05:30:04 |     force_fp16_tokens: False
05:30:04 |     fp16: False
05:30:04 |     fp16_impl: safe
05:30:04 |     fromfile_datapath: /content/drive/MyDrive/sam.txt
05:30:04 |     fromfile_datatype_extension: False
05:30:04 |     gpu: -1
05:30:04 |     gpu_beam_blocking: False
05:30:04 |     gradient_clip: 0.1
05:30:04 |     hide_labels: False
05:30:04 |     history_add_global_end_token: None
05:30:04 |     history_reversed: False
05:30:04 |     history_size: -1
05:30:04 |     image_cropsize: 224
05:30:04 |     image_mode: no_image_model
05:30:04 |     image_size: 256
05:30:04 |     inference: greedy
05:30:04 |     init_model: None
05:30:04 |     init_opt: None
05:30:04 |     interactive_mode: False
05:30:04 |     invsqrt_lr_decay_gamma: -1
05:30:04 |     is_debug: False
05:30:04 |     label_truncate: None
05:30:04 |     learn_positional_embeddings: False
05:30:04 |     learningrate: 1
05:30:04 |     load_from_checkpoint: True
05:30:04 |     log_every_n_secs: -1
05:30:04 |     log_every_n_steps: 50
05:30:04 |     log_keep_fields: all
05:30:04 |     loglevel: info
05:30:04 |     lr_scheduler: reduceonplateau
05:30:04 |     lr_scheduler_decay: 0.5
05:30:04 |     lr_scheduler_patience: 3
05:30:04 |     max_train_steps: -1
05:30:04 |     max_train_time: -1
05:30:04 |     metrics: default
05:30:04 |     model: transformer/generator
05:30:04 |     model_file: /content/drive/MyDrive/sammodel
05:30:04 |     model_parallel: False
05:30:04 |     momentum: 0
05:30:04 |     multitask_weights: [1]
05:30:04 |     mutators: None
05:30:04 |     n_decoder_layers: -1
05:30:04 |     n_encoder_layers: -1
05:30:04 |     n_heads: 2
05:30:04 |     n_layers: 2
05:30:04 |     n_positions: None
05:30:04 |     n_segments: 0
05:30:04 |     nesterov: True
05:30:04 |     no_cuda: False
05:30:04 |     num_epochs: 10.0
05:30:04 |     num_workers: 0
05:30:04 |     nus: (0.7,)
05:30:04 |     optimizer: sgd
05:30:04 |     output_scaling: 1.0
05:30:04 |     override: "{'task': 'fromfile', 'fromfile_datapath': '/content/drive/MyDrive/sam.txt', 'model': 'transformer/generator', 'model_file': '/content/drive/MyDrive/sammodel', 'batchsize': 8, 'num_epochs': 10.0}"
05:30:04 |     parlai_home: /usr/local/lib/python3.10/dist-packages
05:30:04 |     person_tokens: False
05:30:04 |     rank_candidates: False
05:30:04 |     relu_dropout: 0.0
05:30:04 |     save_after_valid: False
05:30:04 |     save_every_n_secs: -1
05:30:04 |     save_format: conversations
05:30:04 |     seed: None
05:30:04 |     share_word_embeddings: True
05:30:04 |     short_final_eval: False
05:30:04 |     skip_generation: False
05:30:04 |     special_tok_lst: None
05:30:04 |     split_lines: False
05:30:04 |     starttime: May03_05-30
05:30:04 |     task: fromfile
05:30:04 |     teacher_seed: None
05:30:04 |     temperature: 1.0
05:30:04 |     tensorboard_log: False
05:30:04 |     tensorboard_logdir: None
05:30:04 |     text_truncate: None
05:30:04 |     topk: 10
05:30:04 |     topp: 0.9
05:30:04 |     truncate: -1
05:30:04 |     update_freq: 1
05:30:04 |     use_reply: label
05:30:04 |     validation_cutoff: 1.0
05:30:04 |     validation_every_n_epochs: -1
05:30:04 |     validation_every_n_secs: -1
05:30:04 |     validation_every_n_steps: -1
05:30:04 |     validation_max_exs: -1
05:30:04 |     validation_metric: accuracy
05:30:04 |     validation_metric_mode: None
05:30:04 |     validation_patience: 10
05:30:04 |     validation_share_agent: False
05:30:04 |     variant: aiayn
05:30:04 |     verbose: False
05:30:04 |     wandb_entity: None
05:30:04 |     wandb_log: False
05:30:04 |     wandb_log_model: False
05:30:04 |     wandb_name: None
05:30:04 |     wandb_project: None
05:30:04 |     warmup_rate: 0.0001
05:30:04 |     warmup_updates: -1
05:30:04 |     weight_decay: None
05:30:04 |     world_logs: 
05:30:04 | creating task(s): fromfile
05:30:04 | Loading ParlAI text data: /content/drive/MyDrive/sam.txt
Traceback (most recent call last):
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1397, in _create_task_agents
    task_agents = my_module.create_agents(opt)  # type: ignore
AttributeError: module 'parlai.tasks.fromfile.agents' has no attribute 'create_agents'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/bin/parlai", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/parlai/__main__.py", line 14, in main
    superscript_main()
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/script.py", line 325, in superscript_main
    return SCRIPT_REGISTRY[cmd].klass._run_from_parser_and_opt(opt, parser)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/script.py", line 108, in _run_from_parser_and_opt
    return script.run()
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/train_model.py", line 1057, in run
    self.train_loop = TrainLoop(self.opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/train_model.py", line 376, in __init__
    build_dict(opt, skip_if_built=True)
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/build_dict.py", line 113, in build_dict
    world_dict = create_task(ordered_opt, dictionary)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1448, in create_task
    world = create_task_world(opt, user_agents, default_world=default_world)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1412, in create_task_world
    task_agents = _create_task_agents(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1400, in _create_task_agents
    return create_task_agent_from_taskname(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/teachers.py", line 2669, in create_task_agent_from_taskname
    task_agents = teacher_class(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/tasks/fromfile/agents.py", line 138, in __init__
    super().__init__(opt, shared)
  File "/usr/local/lib/python3.10/dist-packages/parlai/tasks/fromfile/agents.py", line 107, in __init__
    self._setup_data(datafile)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/teachers.py", line 1578, in _setup_data
    raise ValueError(
ValueError: ParlaiDialogTeacher requires a "labels" field in every entry, but one is missing in Line 1 in /content/drive/MyDrive/sam.txt. The line is:
	text:hello how are you today?   labels:i'm great thanks! what are you doing?

[klshuster]

Huh, could you share perhaps a small sample of your txt file? even the first 2 or 3 lines would be good. There may be a formatting issue

[mohamedae17]

text:Hey, have you seen the new Samsung Galaxy S22? I'm thinking of getting it. labels:No, I haven't. What's so great about it?
text:Well, it has a 6.6-inch AMOLED display with a 120Hz refresh rate. That means everything looks super smooth. labels:Wow, that sounds impressive. How's the battery life?

[krisstud]

text:Hey, have you seen the new Samsung Galaxy S22? I'm thinking of getting it. labels:No, I haven't. What's so great about it?
text:Well, it has a 6.6-inch AMOLED display with a 120Hz refresh rate. That means everything looks super smooth. labels:Wow, that sounds impressive. How's the battery life?

Is there a TAB or space character before "labels:"? It should be:
text:sometexthereTABlabels:more text

[mohamedae17]

there is a TAB
and i has tried another file with this data in documentation :

text:hello how are you today? labels:i'm great thanks! what are you doing?
text:i've just been biking. labels:oh nice, i haven't got on a bike in years! episode_done:True

text:sometextTABlabels:more text
text:sometextTABlabels:more textTABepisode_done:True

and still same error:
07:13:56 | Loading ParlAI text data: /content/drive/MyDrive/sam.txt
Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1397, in _create_task_agents
task_agents = my_module.create_agents(opt) # type: ignore
AttributeError: module 'parlai.tasks.fromfile.agents' has no attribute 'create_agents'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/bin/parlai", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/parlai/__main__.py", line 14, in main
    superscript_main()
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/script.py", line 325, in superscript_main
    return SCRIPT_REGISTRY[cmd].klass._run_from_parser_and_opt(opt, parser)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/script.py", line 108, in _run_from_parser_and_opt
    return script.run()
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/train_model.py", line 1057, in run
    self.train_loop = TrainLoop(self.opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/train_model.py", line 376, in __init__
    build_dict(opt, skip_if_built=True)
  File "/usr/local/lib/python3.10/dist-packages/parlai/scripts/build_dict.py", line 113, in build_dict
    world_dict = create_task(ordered_opt, dictionary)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1448, in create_task
    world = create_task_world(opt, user_agents, default_world=default_world)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1412, in create_task_world
    task_agents = _create_task_agents(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/worlds.py", line 1400, in _create_task_agents
    return create_task_agent_from_taskname(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/teachers.py", line 2669, in create_task_agent_from_taskname
    task_agents = teacher_class(opt)
  File "/usr/local/lib/python3.10/dist-packages/parlai/tasks/fromfile/agents.py", line 138, in __init__
    super().__init__(opt, shared)
  File "/usr/local/lib/python3.10/dist-packages/parlai/tasks/fromfile/agents.py", line 107, in __init__
    self._setup_data(datafile)
  File "/usr/local/lib/python3.10/dist-packages/parlai/core/teachers.py", line 1578, in _setup_data
    raise ValueError(
ValueError: ParlaiDialogTeacher requires a "labels" field in every entry, but one is missing in Line 1 in /content/drive/MyDrive/sam.txt. The line is:
	text:hello how are you today? labels:i'm great thanks! what are you doing?

[mohamedae17]

Oh. it's run now
Editor on colab for text file don't do the same tab
i delete file from drive and create a new txt file on desktop with normal tab and upload it in drive
and finally it's work now
with NEW PROBLEM :"D

[mohamedae17]

after I run this line
!parlai interactive -m transformer/generator -mf /content/drive/MyDrive/sammodel
the output when i talk to chat for testing data it's repeat one word as an answer

Enter [DONE] if you want to end the episode, [EXIT] to quit.
07:36:48 | creating task(s): interactive
Enter Your Message: hello
[TransformerGenerator]: i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i
Enter Your Message: hello how are you today?
[TransformerGenerator]: biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking biking ...
biking biking biking biking biking biking biking biking

[klshuster]

You are training a model from scratch, so this will be significantly worse than models that are initialized with pre-trained model weights

[mohamedae17]

what should i do if i want to getting a chatbot talking about mental health ?
can i add more data for pre-trained model ? or there are ways for making training model from scratch better ?

[klshuster]

yes, you can fine-tune pre-trained models

see e.g. fine-tuning blenderbot 1: https://parl.ai/projects/recipes/