Skip to content
This repository has been archived by the owner on Nov 3, 2023. It is now read-only.

RuntimeError on BB2 during interactivce_web mode with blended_skill_talk task #4121

Closed
samwit opened this issue Oct 28, 2021 · 0 comments · Fixed by #4140
Closed

RuntimeError on BB2 during interactivce_web mode with blended_skill_talk task #4121

samwit opened this issue Oct 28, 2021 · 0 comments · Fixed by #4140

Comments

@samwit
Copy link

samwit commented Oct 28, 2021

Bug description
RuntimeError: There must be exactly two agents for this world.
Getting 'RuntimeError' error when trying to run interactive_web with BB2 and --task blended_skill_talk

Reproduction steps

python -m parlai.scripts.interactive_web --host 0.0.0.0 --task blended_skill_talk --model-file zoo:blenderbot2/blenderbot2_400M/model --search_server 'http://36.135.119.125:1111'

Logs

4:14:22 WARN | Overriding opt["task"] to blended_skill_talk (previously: None)
14:14:22 WARN | Overriding opt["model_file"] to /home/sam/ParlAI/data/models/blenderbot2/blenderbot2_400M/model (previously: /checkpoint/kshust
er/projects/knowledge_bot/kbot_memfix_sweep25_Fri_Jul__9/338/model.oss)
14:14:22 WARN | Overriding opt["search_server"] to http://36.135.119.125:1111 (previously: None)
14:14:22 INFO | loading dictionary from /home/sam/ParlAI/data/models/blenderbot2/blenderbot2_400M/model.dict
14:14:22 INFO | num words = 50264
14:14:22 INFO | BlenderBot2Fid: full interactive mode on.
14:14:34 INFO | Creating the search engine retriever.
Some weights of the model checkpoint at bert-base-uncased were not used when initializing BertModel: ['cls.predictions.transform.dense.weight', `
 'cls.seq_relationship.bias', 'cls.seq_relationship.weight', 'cls.predictions.bias', 'cls.predictions.decoder.weight', 'cls.predictions.transfo
rm.LayerNorm.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.LayerNorm.weight']
- This IS expected if you are initializing BertModel from the checkpoint of a model trained on another task or with another architecture (e.g. 
initializing a BertForSequenceClassification model from a BertForPreTraining model).
- This IS NOT expected if you are initializing BertModel from the checkpoint of a model that you expect to be exactly identical (initializing a
 BertForSequenceClassification model from a BertForSequenceClassification model).
14:14:44 INFO | Building Query Generator from file: /home/sam/ParlAI/data/models/blenderbot2/query_generator/model
14:14:50 INFO | Building Memory Decoder from file: /home/sam/ParlAI/data/models/blenderbot2/memory_decoder/model
14:14:57 INFO | Total parameters: 732,961,280 (731,781,632 trainable)
14:14:57 INFO | Loading existing model params from /home/sam/ParlAI/data/models/blenderbot2/blenderbot2_400M/model
14:14:58 INFO | Opt:
14:14:58 INFO |     activation: gelu
14:14:58 INFO |     adafactor_eps: '[1e-30, 0.001]'
14:14:58 INFO |     adam_eps: 1e-08
14:14:58 INFO |     add_p1_after_newln: False
14:14:58 INFO |     allow_missing_init_opts: False
14:14:58 INFO |     attention_dropout: 0.1
14:14:58 INFO |     batchsize: 12
14:14:58 INFO |     beam_block_full_context: False
14:14:58 INFO |     beam_block_list_filename: None
14:14:58 INFO |     beam_block_ngram: 3
14:14:58 INFO |     beam_context_block_ngram: 3
14:14:58 INFO |     beam_delay: 30
14:14:58 INFO |     beam_length_penalty: 0.65
14:14:58 INFO |     beam_min_length: 20
14:14:58 INFO |     beam_size: 10
14:14:58 INFO |     betas: '[0.9, 0.999]'
14:14:58 INFO |     bpe_add_prefix_space: None
14:14:58 INFO |     bpe_debug: False
14:14:58 INFO |     bpe_dropout: None
14:14:58 INFO |     bpe_merge: None
14:14:58 INFO |     bpe_vocab: None
14:14:58 INFO |     candidates: inline
14:14:58 INFO |     cap_num_predictions: 100
14:14:58 INFO |     checkpoint_activations: False
14:14:58 INFO |     codes_attention_num_heads: 4
14:14:58 INFO |     codes_attention_type: basic
14:14:58 INFO |     compressed_indexer_factory: IVF4096_HNSW128,PQ128
14:14:58 INFO |     compressed_indexer_gpu_train: False
14:14:58 INFO |     compressed_indexer_nprobe: 64
14:14:58 INFO |     compute_tokenized_bleu: False
14:14:58 INFO |     converting: False
14:14:58 INFO |     data_parallel: False
14:14:58 INFO |     datapath: /home/sam/ParlAI/data
14:14:58 INFO |     datatype: train:stream
14:14:58 INFO |     delimiter: '\n'
14:14:58 INFO |     dict_class: parlai.core.dict:DictionaryAgent
14:14:58 INFO |     dict_endtoken: __end__
14:14:58 INFO |     dict_file: /home/sam/ParlAI/data/models/blenderbot2/blenderbot2_400M/model.dict
14:14:58 INFO |     dict_initpath: None
14:14:58 INFO |     dict_language: english
14:14:58 INFO |     dict_loaded: True
14:14:58 INFO |     dict_lower: False
14:14:58 INFO |     dict_max_ngram_size: -1
14:14:58 INFO |     dict_maxtokens: -1
14:14:58 INFO |     dict_minfreq: 0
14:14:58 INFO |     dict_nulltoken: __null__
14:14:58 INFO |     dict_starttoken: __start__
14:14:58 INFO |     dict_textfields: text,labels
14:14:58 INFO |     embedding_projection: random
14:14:58 INFO |     embedding_size: 1024
14:14:58 INFO |     embedding_type: random
14:14:58 INFO |     embeddings_scale: True
14:14:58 INFO |     encode_candidate_vecs: True
14:14:58 INFO |     encode_candidate_vecs_batchsize: 256
14:14:58 INFO |     eval_candidates: inline
14:14:58 INFO |     ffn_size: 4096
14:14:58 INFO |     fixed_candidate_vecs: reuse
14:14:58 INFO |     fixed_candidates_path: None
14:14:58 INFO |     force_fp16_tokens: True
14:14:58 INFO |     fp16: False
14:14:58 INFO |     fp16_impl: safe
14:14:58 INFO |     generation_model: bart
14:14:58 INFO |     gold_document_key: __selected-docs__
14:14:58 INFO |     gold_document_titles_key: select-docs-titles
14:14:58 INFO |     gold_knowledge_passage_key: checked_sentence
14:14:58 INFO |     gold_knowledge_title_key: title
14:14:58 INFO |     gold_sentence_key: __selected-sentences__
14:14:58 INFO |     gpu: -1
14:14:58 INFO |     gradient_clip: 0.1
14:14:58 INFO |     hide_labels: False
14:14:58 INFO |     history_add_global_end_token: None
14:14:58 INFO |     history_reversed: False
14:14:58 INFO |     history_size: -1
14:14:58 INFO |     hnsw_ef_construction: 200
14:14:58 INFO |     hnsw_ef_search: 128
14:14:58 INFO |     hnsw_indexer_store_n: 128
14:14:58 INFO |     host: 0.0.0.0
14:14:58 INFO |     ignore_bad_candidates: False
14:14:58 INFO |     image_cropsize: 224
14:14:58 INFO |     image_mode: raw
14:14:58 INFO |     image_size: 256
14:14:58 INFO |     include_initial_utterances: False
14:14:58 INFO |     include_personas: True
14:14:58 INFO |     memory_decoder_delimiter: '\n'
14:14:58 INFO |     memory_decoder_ignore_phrase: persona:
14:14:58 INFO |     memory_decoder_key: full_text
14:14:58 INFO |     memory_decoder_model_file: zoo:blenderbot2/memory_decoder/model
14:14:58 INFO |     memory_decoder_one_line_memories: False
14:14:58 INFO |     memory_decoder_truncate: -1
14:14:58 INFO |     memory_doc_delimiter: :
14:14:58 INFO |     memory_doc_title_delimiter: ' / '
14:14:58 INFO |     memory_extractor_phrase: persona:
14:14:58 INFO |     memory_key: personas
14:14:58 INFO |     memory_reader_model: None
14:14:58 INFO |     memory_retriever_truncate: -1
14:14:58 INFO |     memory_writer_model: bert
14:14:58 INFO |     memory_writer_model_file: zoo:hallucination/multiset_dpr/hf_bert_base.cp
14:14:58 INFO |     min_doc_token_length: 64
14:14:58 INFO |     model: projects.blenderbot2.agents.blenderbot2:BlenderBot2FidAgent
14:14:58 INFO |     model_file: /home/sam/ParlAI/data/models/blenderbot2/blenderbot2_400M/model
14:14:58 INFO |     model_parallel: True
14:14:58 INFO |     momentum: 0
14:14:58 INFO |     multitask_weights: '[3.0, 1.0, 1.0, 1.0, 3.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]'
14:14:58 INFO |     mutators: None
14:14:58 INFO |     n_decoder_layers: 12
14:14:58 INFO |     n_docs: 5
14:14:58 INFO |     n_encoder_layers: 12
14:14:58 INFO |     n_extra_positions: 0
14:14:58 INFO |     n_heads: 16
14:14:58 INFO |     n_layers: 12
14:14:58 INFO |     n_positions: 1024
14:14:58 INFO |     n_ranked_doc_chunks: 1
14:14:58 INFO |     n_segments: 0
14:14:58 INFO |     nesterov: True
14:14:58 INFO |     no_cuda: False
14:14:58 INFO |     normalize_sent_emb: False
14:14:58 INFO |     nus: [0.7]
14:14:58 INFO |     optimizer: adamax
14:14:58 INFO |     outfile: 
14:14:58 INFO |     output_scaling: 1.0
14:14:58 INFO |     override: "{'host': '0.0.0.0', 'task': 'blended_skill_talk', 'model_file': '/home/sam/ParlAI/data/models/blenderbot2/blende
rbot2_400M/model', 'search_server': 'http://36.135.119.125:1111'}"
14:14:58 INFO |     parlai_home: /private/home/kshuster/ParlAI
14:14:58 INFO |     path_to_dense_embeddings: None
14:14:58 INFO |     path_to_dpr_passages: zoo:hallucination/wiki_passages/psgs_w100.tsv
14:14:58 INFO |     path_to_index: zoo:hallucination/wiki_index_compressed/compressed_pq
14:14:58 INFO |     person_tokens: False
14:14:58 INFO |     poly_attention_num_heads: 4
14:14:58 INFO |     poly_attention_type: basic
14:14:58 INFO |     poly_faiss_model_file: None
14:14:58 INFO |     poly_n_codes: 64
14:14:58 INFO |     poly_score_initial_lambda: 0.5
14:14:58 INFO |     polyencoder_init_model: wikito
14:14:58 INFO |     polyencoder_type: codes
14:14:58 INFO |     port: 8080
14:14:58 INFO |     print_docs: False
14:14:58 INFO |     query_generator_beam_min_length: 2
14:14:58 INFO |     query_generator_beam_size: 1
14:14:58 INFO |     query_generator_delimiter: '\n'
14:14:58 INFO |     query_generator_ignore_phrase: persona:
14:14:58 INFO |     query_generator_inference: beam
14:14:58 INFO |     query_generator_key: full_text
14:14:58 INFO |     query_generator_model_file: zoo:blenderbot2/query_generator/model
14:14:58 INFO |     query_generator_truncate: -1
14:14:58 INFO |     query_model: bert_from_parlai_rag
14:14:58 INFO |     rag_model_type: token
14:14:58 INFO |     rag_query_truncate: 512
14:14:58 INFO |     rag_retriever_query: full_history
14:14:58 INFO |     rag_retriever_type: search_engine
14:14:58 INFO |     rag_turn_discount_factor: 1.0
14:14:58 INFO |     rag_turn_marginalize: doc_then_turn
14:14:58 INFO |     retriever_delimiter: '\n'
14:14:58 INFO |     retriever_embedding_size: 768
14:14:58 INFO |     retriever_ignore_phrase: persona:
14:14:58 INFO |     return_cand_scores: False
14:14:58 INFO |     safe_personas_only: True
14:14:58 INFO |     save_format: conversations
14:14:58 INFO |     search_query_generator_beam_min_length: 2
14:14:58 INFO |     search_query_generator_beam_size: 1
14:14:58 INFO |     search_query_generator_inference: greedy
14:14:58 INFO |     search_query_generator_model_file: zoo:blenderbot2/query_generator/model
14:14:58 INFO |     search_query_generator_text_truncate: 512
14:14:58 INFO |     search_server: http://36.135.119.125:1111
14:14:58 INFO |     share_encoders: True
14:14:58 INFO |     share_search_and_memory_query_encoder: False
14:14:58 INFO |     share_word_embeddings: True
14:14:58 INFO |     single_turn: False
14:14:58 INFO |     skip_generation: False
14:14:58 INFO |     skip_retrieval_token: no_passages_used
14:14:58 INFO |     special_tok_lst: None
14:14:58 INFO |     split_lines: True
14:14:58 INFO |     splitted_chunk_length: 256
14:14:58 INFO |     starttime: Jul09_14-09
14:14:58 INFO |     t5_dropout: 0.0
14:14:58 INFO |     t5_generation_config: None
14:14:58 INFO |     t5_model_arch: t5-base
14:14:58 INFO |     t5_model_parallel: False
14:14:58 INFO |     task: blended_skill_talk
14:14:58 INFO |     temperature: 1.0
14:14:58 INFO |     text_truncate: 512
14:14:58 INFO |     tfidf_max_doc_paragraphs: -1
14:14:58 INFO |     tfidf_model_path: zoo:wikipedia_full/tfidf_retriever/model
14:14:58 INFO |     thorough: False
14:14:58 INFO |     topk: 10
14:14:58 INFO |     topp: 0.9
14:14:58 INFO |     train_predict: False
14:14:58 INFO | Current ParlAI commit: 33ce8b5b4cd40d93f22da097d5eb6ce475c56ca8
14:14:58 INFO | creating task(s): blended_skill_talk

Traceback (most recent call last):
  File "/opt/conda/lib/python3.7/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/opt/conda/lib/python3.7/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/sam/ParlAI/parlai/scripts/interactive_web.py", line 302, in <module>
    InteractiveWeb.main()
  File "/home/sam/ParlAI/parlai/core/script.py", line 129, in main
    return cls._run_args(None)
  File "/home/sam/ParlAI/parlai/core/script.py", line 101, in _run_args
    return cls._run_from_parser_and_opt(opt, parser)
  File "/home/sam/ParlAI/parlai/core/script.py", line 108, in _run_from_parser_and_opt
    return script.run()
  File "/home/sam/ParlAI/parlai/scripts/interactive_web.py", line 298, in run
    return interactive_web(self.opt)
  File "/home/sam/ParlAI/parlai/scripts/interactive_web.py", line 276, in interactive_web
    SHARED['world'] = create_task(SHARED.get('opt'), SHARED['agent'])
  File "/home/sam/ParlAI/parlai/core/worlds.py", line 1441, in create_task
    world = create_task_world(opt, user_agents, default_world=default_world)
  File "/home/sam/ParlAI/parlai/core/worlds.py", line 1414, in create_task_world
    return world_class(opt, task_agents + user_agents)
  File "/home/sam/ParlAI/parlai/tasks/blended_skill_talk/worlds.py", line 139, in __init__
    super().__init__(opt, agents, shared)
  File "/home/sam/ParlAI/parlai/tasks/interactive/worlds.py", line 34, in __init__
    super().__init__(opt, agents, shared)
  File "/home/sam/ParlAI/parlai/core/worlds.py", line 340, in __init__
    raise RuntimeError('There must be exactly two agents for this world.')
RuntimeError: There must be exactly two agents for this world.

Additional context
Running with safe interactive mode is working fine

python -m parlai.scripts.safe_interactive --task blended_skill_talk --model-file zoo:blender
bot2/blenderbot2_400M/model --search_server 'http://36.135.119.125:1111'
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Fix interactive web facebookresearch/ParlAI
1 participant
@samwit