ner.correct recipe : "No tasks available" after checking few instances

[fsa]

I am using the ner.correct recipe:

prodigy ner.correct training_data /model/ /prodigy_jsonl/demo_jsonl.jsonl --label "A"

to suggest to the user 300 training instances annotated using a temporary trained Spacy model. The user should interact with prodigy interface and confirm or correct the annotations. However, after checking few instances (mostly 10 instances) the interface stopped and showed No tasks available. I created also fresh database but always same issue. The json format is correct for all instances also there is no duplicate. Each instance is almost 4 or 5 lines long.

here is the prodigy.json:

 {
  "theme": "basic",
  "custom_theme": {},
  "buttons": ["accept", "reject", "ignore", "undo"],
  "custom_theme": {
     "cardMaxWidth":1200
    },
  "batch_size": 10,
  "history_size": 10,
  "port": 7777,
  "host": "xxxxx",
  "cors": true,
  "db": "sqlite",
  "db_settings": {},
  "api_keys": {},
  "validate": true,
  "auto_exclude_current": true,
  "instant_submit": false,
  "feed_overlap": false,
  "ui_lang": "en",
  "project_info": ["dataset", "session", "lang", "recipe_name", "view_id", "label"],
  "show_stats": false,
  "hide_meta": false,
  "show_flag": false,
  "instructions": false,
  "swipe": false,
  "swipe_gestures": { "left": "accept", "right": "reject" },
  "split_sents_threshold": false,
  "global_css": null,
  "javascript": null,
  "writing_dir": "ltr",
  "show_whitespace": false
}

[ryanwesslen]

hi @fsa!

It's tough to say as it doesn't look like anything obvious. Have you seen this post?

No tasks available for ner.correct?

Hi! How large are your individual texts in the data? And are you loading from a format that can be read line by line (e.g. JSONL) or a format that needs to be parsed and loaded into memory upfront (e.g. JSON)? "No tasks available" is shown if Prodigy comes across an empty batch and typically happens when there's nothing more to annotate. If there is more data, this could indicate that the server somehow takes too long to produce a new batch and/or some race conditon causes an empty batch to be …

Do you have any duplicates? I'm wondering if hashing could be a problem.

No tasks available with ner.correct and prodigy 1.11.0

Hi! So just to confirm, the remaining examples in your dataset are all examples with an _input_hash that's not yet in the current dataset you're saving your annotations to? One simple way to check this would be to load your input examples, call prodigy.set_hashes() on each of them and then check the _input_hash values against the list of input hashes in the dataset.

I've found several other related posts:

Prodigy says "No tasks available."

I'm using the mark recipe to add NER annotations to a dataset of 1000 documents. After just 10 documents, I get "No tasks available." and I cannot do anything other than close the application. When I restart, all my annotations are gone (even though I clicked the save button), and I'm shown the same 10 documents. I would like to be able to view and annotate the entire dataset; how can I do that? I'm using prodigy 1.10.8.

Have you looked through these as well?

Last, have you worked with logging? Does it provide any insight?

Let me know if after these posts if you're still not successful. Unfortunately without the exact data it's hard to recreate on our end.

[fsa]

Hi Ryan

I read all of these posts but unfortunately still no solution. As I mentioned in my post there is no duplicate and the format is correct.
When I take only the first 28 instances out of the 300 it works. Before with loading the 300 instances after 10 instances No tasks available appears. I will check the data more deeper and report here what will happened.

[fsa]

Using fresh Prodigy server instance solved the issue. So far I could control the annotations for 1000 instances without any issue.