Introduction filename correction (#1239)

* intro file name fix

guides/README.md

@@ -2,7 +2,7 @@
 
 Also check out our [Examples](../examples) to see `jiant` in action.
 
-If you don't know what to read, why not read our [In-Depth Introduction to Jiant](general/in_depth_into.md)?
+If you don't know what to read, why not read our [In-Depth Introduction to Jiant](general/in_depth_intro.md)?
 
 Contents:
 
@@ -17,24 +17,24 @@ Contents:
 
 ## Tutorials
 
-These are quick tutorials that demonstrate `jiant` usage. 
+These are quick tutorials that demonstrate `jiant` usage.
 
 * [Quick Start Guide — Using the "Simple" CLI](tutorials/quick_start_simple.md): A simple `jiant` training run in bash, using the "Simple" CLI
 * [Quick Start Guide — Using the "Main" CLI](tutorials/quick_start_main.md): A simple `jiant` training run in bash, using the "Main" CLI
 
-The "Simple" API provides a single command-line script for training and evaluating models on tasks, while the "Main" API offers more flexibilty by breaking the workflow down into discrete steps (downloading the model, tokenization & caching, writing a fully specific run-configuration, and finally running the experiment). Both interfaces use the same models and task implementations uner the hood. 
+The "Simple" API provides a single command-line script for training and evaluating models on tasks, while the "Main" API offers more flexibilty by breaking the workflow down into discrete steps (downloading the model, tokenization & caching, writing a fully specific run-configuration, and finally running the experiment). Both interfaces use the same models and task implementations uner the hood.
 
 
 ## General
 
 These are general guides to `jiant`'s design and components. Refer to these if you have questions about parts of `jiant`:
 
-* [In-Depth Introduction to Jiant](general/in_depth_into.md): Learn about `jiant` in greater detail
-    * [`jiant`'s models](general/in_depth_into.md#jiants-models)
-    * [`jiant`'s tasks](general/in_depth_into.md#jiants-tasks)
-    * [`Runner`s and `Metarunner`s](general/in_depth_into.md#runners-and-metarunners)
-    * [Step-by-step through `jiant`'s pipeline](general/in_depth_into.md#step-by-step-through-jiants-pipeline)
-    
+* [In-Depth Introduction to Jiant](general/in_depth_intro.md): Learn about `jiant` in greater detail
+    * [`jiant`'s models](general/in_depth_intro.md#jiants-models)
+    * [`jiant`'s tasks](general/in_depth_intro.md#jiants-tasks)
+    * [`Runner`s and `Metarunner`s](general/in_depth_intro.md#runners-and-metarunners)
+    * [Step-by-step through `jiant`'s pipeline](general/in_depth_intro.md#step-by-step-through-jiants-pipeline)
+
 ## Running benchmarks
 
 These are guides to running common NLP benchmarks using `jiant`:
@@ -56,7 +56,7 @@ These are notes on the tasks supported in `jiant`:
 
 * [List of supported tasks in `jiant`](tasks/supported_tasks.md)
 * [Task-specific notes](tasks/task_specific.md): Learn about quirks/caveats about specific tasks
-* [Adding Tasks](tasks/adding_tasks.md): Guide on adding a task to `jiant` 
+* [Adding Tasks](tasks/adding_tasks.md): Guide on adding a task to `jiant`
 
 ## Papers / Projects
 

guides/tutorials/quick_start_main.md

@@ -14,7 +14,7 @@ TASK=rte
 ```
 
 Neither of these environment variables are read directly in Python, but simply used to make our bash commands more succinct. This also writes the intermediate and output files in our recommended folder organization.
- 
+
 1. We'll get the data using `jiant`'s download script
 ```bash
 python jiant/scripts/download_data/runscript.py \
@@ -41,9 +41,9 @@ python jiant/proj/main/tokenize_and_cache.py \
     --phases train,val \
     --max_seq_length 256 \
     --smart_truncate
-```  
+```
 
-4. Next, we write a [run-config](../general/in_depth_into.md#write-run-config). This writes a JSON file that specifies some configuration for our run - it's pretty simple now, but will be much more meaningful when running more complex multitasking experiments. 
+4. Next, we write a [run-config](../general/in_depth_intro.md#write-run-config). This writes a JSON file that specifies some configuration for our run - it's pretty simple now, but will be much more meaningful when running more complex multitasking experiments.
 ```bash
 python jiant/proj/main/scripts/configurator.py \
     SingleTaskConfigurator \
@@ -95,12 +95,15 @@ python jiant/proj/main/runscript.py \
         ...
         ...
     --save_checkpoint_every_steps 500 \
+    --delete_checkpoint_if_done
 ```
 
 This will save a checkpoint to disk every 500 training steps. The checkpoint will be saved to a `checkpoint.p` file. If the process gets killed, you can rerun the exact same command and it will continue training from the latest checkpoint.
 
 Note that checkpoints are for resuming training, not for saving snapshots of model weights at different points in training. Checkpoints also include additional run metadata, as well as the optimizer states. To save regular snapshots of model weights, see [Model Snapshots](#model-snapshots)
 
+We also set the `delete_checkpoint_if_done` flag to delete the checkpoint after training is complete.
+
 ### Model Snapshots
 
 To save snapshots of model weights at regular intervals, use the `--save_every_steps` argument. For example:
@@ -113,7 +116,7 @@ will save a pickle of model weights every 500 training steps.
 
 ### Early Stopping
 
-To do early stopping, we can perform validation evaluation at regular intervals over the course of training, and select the best model weights based on validation performance. For expedience, we often do not want to evaluate on the whole validation set, but only a subset. To do early stopping, use the following arguments as an example: 
+To do early stopping, we can perform validation evaluation at regular intervals over the course of training, and select the best model weights based on validation performance. For expedience, we often do not want to evaluate on the whole validation set, but only a subset. To do early stopping, use the following arguments as an example:
 
 ```
     --eval_every_steps 1000

guides/tutorials/quick_start_simple.md

@@ -11,7 +11,7 @@ First, let us assume that we will be working with the following working director
 EXP_DIR=/path/to/exp
 ```
 
-For this training example we'll use the RTE task from GLUE, and the RoBERTa-base model. 
+For this training example we'll use the RTE task from GLUE, and the RoBERTa-base model.
 
 1. We'll get the data using `jiant`'s download script
 ```bash
@@ -36,7 +36,7 @@ python jiant/proj/simple/runscript.py \
     --do_save
 ```
 
-The "Simple" CLI subsumes several steps under the hood, including downloading the `roberta-base` model, tokenizing and caching the data, writing a [run-configuration](../general/in_depth_into.md#write-run-config), and performing the training. 
+The "Simple" CLI subsumes several steps under the hood, including downloading the `roberta-base` model, tokenizing and caching the data, writing a [run-configuration](../general/in_depth_intro.md#write-run-config), and performing the training.
 
 
 ## Additional Options
@@ -51,7 +51,7 @@ python jiant/proj/simple/runscript.py \
     --do_save
 ```
 
-This will save two sets of model weights: `last_model.p` will the the model weights at the end of training, while `best_model.p` will be the weights which achieves the best validation score on the evaluation subset (see: [Early Stopping](#early-stopping)). If you only one to save one or the other, use `--do_save_last` or `--do_save_best`. 
+This will save two sets of model weights: `last_model.p` will the the model weights at the end of training, while `best_model.p` will be the weights which achieves the best validation score on the evaluation subset (see: [Early Stopping](#early-stopping)). If you only one to save one or the other, use `--do_save_last` or `--do_save_best`.
 
 ### Checkpointing
 
@@ -70,7 +70,7 @@ This will save a checkpoint to disk every 500 training steps. The checkpoint wil
 
 Note that checkpoints are for resuming training, not for saving snapshots of model weights at different points in training. Checkpoints also include additional run metadata, as well as the optimizer states. To save regular snapshots of model weights, see [Model Snapshots](#model-snapshots)
 
-We also set the `delete_checkpoint_if_done` flag to delete the checkpoint after training is complete. 
+We also set the `delete_checkpoint_if_done` flag to delete the checkpoint after training is complete.
 
 ### Model Snapshots
 
@@ -84,7 +84,7 @@ will save a pickle of model weights every 500 training steps.
 
 ### Early Stopping
 
-To do early stopping, we can perform validation evaluation at regular intervals over the course of training, and select the best model weights based on validation performance. For expedience, we often do not want to evaluate on the whole validation set, but only a subset. To do early stopping, use the following arguments as an example: 
+To do early stopping, we can perform validation evaluation at regular intervals over the course of training, and select the best model weights based on validation performance. For expedience, we often do not want to evaluate on the whole validation set, but only a subset. To do early stopping, use the following arguments as an example:
 
 ```
     --eval_every_steps 1000