InvokeAI/docs/contributing/DOWNLOAD_QUEUE.md

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

278 lines
12 KiB
Markdown
Raw Normal View History

[feature] Download Queue (#5225) * add base definition of download manager * basic functionality working * add unit tests for download queue * add documentation and FastAPI route * fix docs * add missing test dependency; fix import ordering * fix file path length checking on windows * fix ruff check error * move release() into the __del__ method * disable testing of stderr messages due to issues with pytest capsys fixture * fix unsorted imports * harmonized implementation of start() and stop() calls in download and & install modules * Update invokeai/app/services/download/download_base.py Co-authored-by: Ryan Dick <ryanjdick3@gmail.com> * replace test datadir fixture with tmp_path * replace DownloadJobBase->DownloadJob in download manager documentation * make source and dest arguments to download_queue.download() an AnyHttpURL and Path respectively * fix pydantic typecheck errors in the download unit test * ruff formatting * add "job cancelled" as an event rather than an exception * fix ruff errors * Update invokeai/app/services/download/download_default.py Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com> * use threading.Event to stop service worker threads; handle unfinished job edge cases * remove dangling STOP job definition * fix ruff complaint * fix ruff check again * avoid race condition when start() and stop() are called simultaneously from different threads * avoid race condition in stop() when a job becomes active while shutting down --------- Co-authored-by: Lincoln Stein <lstein@gmail.com> Co-authored-by: Ryan Dick <ryanjdick3@gmail.com> Co-authored-by: psychedelicious <4822129+psychedelicious@users.noreply.github.com> Co-authored-by: Kent Keirsey <31807370+hipsterusername@users.noreply.github.com>
2023-12-22 17:35:57 +00:00
# The InvokeAI Download Queue
The DownloadQueueService provides a multithreaded parallel download
queue for arbitrary URLs, with queue prioritization, event handling,
and restart capabilities.
## Simple Example
```
from invokeai.app.services.download import DownloadQueueService, TqdmProgress
download_queue = DownloadQueueService()
for url in ['https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/a-painting-of-a-fire.png?raw=true',
'https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/birdhouse.png?raw=true',
'https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/missing.png',
'https://civitai.com/api/download/models/152309?type=Model&format=SafeTensor',
]:
# urls start downloading as soon as download() is called
download_queue.download(source=url,
dest='/tmp/downloads',
on_progress=TqdmProgress().update
)
download_queue.join() # wait for all downloads to finish
for job in download_queue.list_jobs():
print(job.model_dump_json(exclude_none=True, indent=4),"\n")
```
Output:
```
{
"source": "https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/a-painting-of-a-fire.png?raw=true",
"dest": "/tmp/downloads",
"id": 0,
"priority": 10,
"status": "completed",
"download_path": "/tmp/downloads/a-painting-of-a-fire.png",
"job_started": "2023-12-04T05:34:41.742174",
"job_ended": "2023-12-04T05:34:42.592035",
"bytes": 666734,
"total_bytes": 666734
}
{
"source": "https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/birdhouse.png?raw=true",
"dest": "/tmp/downloads",
"id": 1,
"priority": 10,
"status": "completed",
"download_path": "/tmp/downloads/birdhouse.png",
"job_started": "2023-12-04T05:34:41.741975",
"job_ended": "2023-12-04T05:34:42.652841",
"bytes": 774949,
"total_bytes": 774949
}
{
"source": "https://github.com/invoke-ai/InvokeAI/blob/main/invokeai/assets/missing.png",
"dest": "/tmp/downloads",
"id": 2,
"priority": 10,
"status": "error",
"job_started": "2023-12-04T05:34:41.742079",
"job_ended": "2023-12-04T05:34:42.147625",
"bytes": 0,
"total_bytes": 0,
"error_type": "HTTPError(Not Found)",
"error": "Traceback (most recent call last):\n File \"/home/lstein/Projects/InvokeAI/invokeai/app/services/download/download_default.py\", line 182, in _download_next_item\n self._do_download(job)\n File \"/home/lstein/Projects/InvokeAI/invokeai/app/services/download/download_default.py\", line 206, in _do_download\n raise HTTPError(resp.reason)\nrequests.exceptions.HTTPError: Not Found\n"
}
{
"source": "https://civitai.com/api/download/models/152309?type=Model&format=SafeTensor",
"dest": "/tmp/downloads",
"id": 3,
"priority": 10,
"status": "completed",
"download_path": "/tmp/downloads/xl_more_art-full_v1.safetensors",
"job_started": "2023-12-04T05:34:42.147645",
"job_ended": "2023-12-04T05:34:43.735990",
"bytes": 719020768,
"total_bytes": 719020768
}
```
## The API
The default download queue is `DownloadQueueService`, an
implementation of ABC `DownloadQueueServiceBase`. It juggles multiple
background download requests and provides facilities for interrogating
and cancelling the requests. Access to a current or past download task
is mediated via `DownloadJob` objects which report the current status
of a job request
### The Queue Object
A default download queue is located in
`ApiDependencies.invoker.services.download_queue`. However, you can
create additional instances if you need to isolate your queue from the
main one.
```
queue = DownloadQueueService(event_bus=events)
```
`DownloadQueueService()` takes three optional arguments:
| **Argument** | **Type** | **Default** | **Description** |
|----------------|-----------------|---------------|-----------------|
| `max_parallel_dl` | int | 5 | Maximum number of simultaneous downloads allowed |
| `event_bus` | EventServiceBase | None | System-wide FastAPI event bus for reporting download events |
| `requests_session` | requests.sessions.Session | None | An alternative requests Session object to use for the download |
`max_parallel_dl` specifies how many download jobs are allowed to run
simultaneously. Each will run in a different thread of execution.
`event_bus` is an EventServiceBase, typically the one created at
InvokeAI startup. If present, download events are periodically emitted
on this bus to allow clients to follow download progress.
`requests_session` is a url library requests Session object. It is
used for testing.
### The Job object
The queue operates on a series of download job objects. These objects
specify the source and destination of the download, and keep track of
the progress of the download.
The only job type currently implemented is `DownloadJob`, a pydantic object with the
following fields:
| **Field** | **Type** | **Default** | **Description** |
|----------------|-----------------|---------------|-----------------|
| _Fields passed in at job creation time_ |
| `source` | AnyHttpUrl | | Where to download from |
| `dest` | Path | | Where to download to |
| `access_token` | str | | [optional] string containing authentication token for access |
| `on_start` | Callable | | [optional] callback when the download starts |
| `on_progress` | Callable | | [optional] callback called at intervals during download progress |
| `on_complete` | Callable | | [optional] callback called after successful download completion |
| `on_error` | Callable | | [optional] callback called after an error occurs |
| `id` | int | auto assigned | Job ID, an integer >= 0 |
| `priority` | int | 10 | Job priority. Lower priorities run before higher priorities |
| |
| _Fields updated over the course of the download task_
| `status` | DownloadJobStatus| | Status code |
| `download_path` | Path | | Path to the location of the downloaded file |
| `job_started` | float | | Timestamp for when the job started running |
| `job_ended` | float | | Timestamp for when the job completed or errored out |
| `job_sequence` | int | | A counter that is incremented each time a model is dequeued |
| `bytes` | int | 0 | Bytes downloaded so far |
| `total_bytes` | int | 0 | Total size of the file at the remote site |
| `error_type` | str | | String version of the exception that caused an error during download |
| `error` | str | | String version of the traceback associated with an error |
| `cancelled` | bool | False | Set to true if the job was cancelled by the caller|
When you create a job, you can assign it a `priority`. If multiple
jobs are queued, the job with the lowest priority runs first.
Every job has a `source` and a `dest`. `source` is a pydantic.networks AnyHttpUrl object.
The `dest` is a path on the local filesystem that specifies the
destination for the downloaded object. Its semantics are
described below.
When the job is submitted, it is assigned a numeric `id`. The id can
then be used to fetch the job object from the queue.
The `status` field is updated by the queue to indicate where the job
is in its lifecycle. Values are defined in the string enum
`DownloadJobStatus`, a symbol available from
`invokeai.app.services.download_manager`. Possible values are:
| **Value** | **String Value** | ** Description ** |
|--------------|---------------------|-------------------|
| `WAITING` | waiting | Job is on the queue but not yet running|
| `RUNNING` | running | The download is started |
| `COMPLETED` | completed | Job has finished its work without an error |
| `ERROR` | error | Job encountered an error and will not run again|
`job_started` and `job_ended` indicate when the job
was started (using a python timestamp) and when it completed.
In case of an error, the job's status will be set to `DownloadJobStatus.ERROR`, the text of the
Exception that caused the error will be placed in the `error_type`
field and the traceback that led to the error will be in `error`.
A cancelled job will have status `DownloadJobStatus.ERROR` and an
`error_type` field of "DownloadJobCancelledException". In addition,
the job's `cancelled` property will be set to True.
### Callbacks
Download jobs can be associated with a series of callbacks, each with
the signature `Callable[["DownloadJob"], None]`. The callbacks are assigned
using optional arguments `on_start`, `on_progress`, `on_complete` and
`on_error`. When the corresponding event occurs, the callback wil be
invoked and passed the job. The callback will be run in a `try:`
context in the same thread as the download job. Any exceptions that
occur during execution of the callback will be caught and converted
into a log error message, thereby allowing the download to continue.
#### `TqdmProgress`
The `invokeai.app.services.download.download_default` module defines a
class named `TqdmProgress` which can be used as an `on_progress`
handler to display a completion bar in the console. Use as follows:
```
from invokeai.app.services.download import TqdmProgress
download_queue.download(source='http://some.server.somewhere/some_file',
dest='/tmp/downloads',
on_progress=TqdmProgress().update
)
```
### Events
If the queue was initialized with the InvokeAI event bus (the case
when using `ApiDependencies.invoker.services.download_queue`), then
download events will also be issued on the bus. The events are:
* `download_started` -- This is issued when a job is taken off the
queue and a request is made to the remote server for the URL headers, but before any data
has been downloaded. The event payload will contain the keys `source`
and `download_path`. The latter contains the path that the URL will be
downloaded to.
* `download_progress -- This is issued periodically as the download
runs. The payload contains the keys `source`, `download_path`,
`current_bytes` and `total_bytes`. The latter two fields can be
used to display the percent complete.
* `download_complete` -- This is issued when the download completes
successfully. The payload contains the keys `source`, `download_path`
and `total_bytes`.
* `download_error` -- This is issued when the download stops because
of an error condition. The payload contains the fields `error_type`
and `error`. The former is the text representation of the exception,
and the latter is a traceback showing where the error occurred.
### Job control
To create a job call the queue's `download()` method. You can list all
jobs using `list_jobs()`, fetch a single job by its with
`id_to_job()`, cancel a running job with `cancel_job()`, cancel all
running jobs with `cancel_all_jobs()`, and wait for all jobs to finish
with `join()`.
#### job = queue.download(source, dest, priority, access_token)
Create a new download job and put it on the queue, returning the
DownloadJob object.
#### jobs = queue.list_jobs()
Return a list of all active and inactive `DownloadJob`s.
#### job = queue.id_to_job(id)
Return the job corresponding to given ID.
Return a list of all active and inactive `DownloadJob`s.
#### queue.prune_jobs()
Remove inactive (complete or errored) jobs from the listing returned
by `list_jobs()`.
#### queue.join()
Block until all pending jobs have run to completion or errored out.