Updating training fits

[alexpiet]

First of possibly many PRs about updating the code to fit training sessions. My first goal is to build the infrastructure for getting a manifest of sessions. Then I will update the fitting scripts. The guts of the model should be fine with the exception of early training sessions, but I will need to review them. Finally I will review the training summary file

• resolves Fit training sessions #92

Manifest and Inventory

• resolves Training - Make manifest to get sessions #325
• resolves Training - do inventory tools #329
• ptt.get_training_manifest() generates a list of all training sessions associated with the mice in summary_df
• ptt.get_training_inventory()

Fitting Scripts

• scripts/deploy_training.sh
• scripts/deploy_training.py
• scripts/fit_training.py
• Removes three old fitting scripts

Model internals

• Should handle no omissions fine, but check
• Check how things like QC and data alignment are handled
  • resolves Training - data QC, alignment, fitting #326
• Early training without flashed images ?
  • Training - pseudo flashes for early stages #327
  • Making a decision to skip these for now

Training Summary file

• ptt.get_training_summary_table() loads the saved training summary file
• ptt.build_training_summary_table() computes and saves the training summary file
• ptt.build_core_training_table() loads the model fits and compiles model fit information
• ptt.add_time_aligned_training_info() saves image by image information to the summary file
• ptt.add_training_engagement_metrics() annotates engagement status.

Analysis

• Resolves Training - do analysis #328
• update old functions
  • ptt.plot_average_by_stage()
  • ptt.plot_average_by_stage_inner()
  • ptt.clean_session_type() is redundant with pgt.get_clean_session_name
  • ptt.get_mouse_pivot_table()
  • ptt.plot_mouse_strategy_correlation()
  • ptt.plot_average_by_day()
• Plot each metric three ways- by stage, by day forward from training start, by day backwards from ophys
• Remove plot_all_ functions and use a list of metrics to call the relevant plot functions
• Use psy_style for colors
• Make the figure sizes better
• It looks like "fraction engaged" decreases with training, but num_rewards is constant. For both strategies! Does this mean the mice are improving their lick hit fraction? worth exploring