- 86932469e76f1315ee18bfa2fc52b588241dace1 add image_to_dataURL util
- 0c2611059711b45bb6142d30b1d1343ac24268f3 make fast latents method
static
- this method doesn't really need `self` and should be able to be called
without instantiating `Generator`
- 2360bfb6558ea511e9c9576f3d4b5535870d84b4 fix schema gen for
GraphExecutionState
- `GraphExecutionState` uses `default_factory` in its fields; the result
is the OpenAPI schema marks those fields as optional, which propagates
to the generated API client, which means we need a lot of unnecessary
type guards to use this data type. the [simple
fix](https://github.com/pydantic/pydantic/discussions/4577) is to add
config to explicitly say all class properties are required. looks this
this will be resolved in a future pydantic release
- 3cd7319cfdb0f07c6bb12d62d7d02efe1ab12675 fix step callback and fast
latent generation on nodes. have this working in UI. depends on the
small change in #2957
Update `compel` to 1.0.0.
This fixes#2832.
It also changes the way downweighting is applied. In particular,
downweighting should now be much better and more controllable.
From the [compel
changelog](https://github.com/damian0815/compel#changelog):
> Downweighting now works by applying an attention mask to remove the
downweighted tokens, rather than literally removing them from the
sequence. This behaviour is the default, but the old behaviour can be
re-enabled by passing `downweight_mode=DownweightMode.REMOVE` on init of
the `Compel` instance.
>
> Formerly, downweighting a token worked by both multiplying the
weighting of the token's embedding, and doing an inverse-weighted blend
with a copy of the token sequence that had the downweighted tokens
removed. The intuition is that as weight approaches zero, the tokens
being downweighted should be actually removed from the sequence.
However, removing the tokens resulted in the positioning of all
downstream tokens becoming messed up. The blend ended up blending a lot
more than just the tokens in question.
>
> As of v1.0.0, taking advice from @keturn and @bonlime
(https://github.com/damian0815/compel/issues/7) the procedure is by
default different. Downweighting still involves a blend but what is
blended is a version of the token sequence with the downweighted tokens
masked out, rather than removed. This correctly preserves positioning
embeddings of the other tokens.
* Update root component to allow optional children that will render as
dynamic header of UI
* Export additional components (logo & themeChanger) for use in said
dynamic header (more to come here)
# The Problem
Pickle files (.pkl, .ckpt, etc) are extremely unsafe as they can be
trivially crafted to execute arbitrary code when parsed using
`torch.load`
Right now the conventional wisdom among ML researchers and users is to
simply `not run untrusted pickle files ever` and instead only use
Safetensor files, which cannot be injected with arbitrary code. This is
very good advice.
Unfortunately, **I have discovered a vulnerability inside of InvokeAI
that allows an attacker to disguise a pickle file as a safetensor and
have the payload execute within InvokeAI.**
# How It Works
Within `model_manager.py` and `convert_ckpt_to_diffusers.py` there are
if-statements that decide which `load` method to use based on the file
extension of the model file. The logic (written in a slightly more
readable format than it exists in the codebase) is as follows:
```
if Path(file).suffix == '.safetensors':
safetensor_load(file)
else:
unsafe_pickle_load(file)
```
A malicious actor would only need to create an infected .ckpt file, and
then rename the extension to something that does not pass the `==
'.safetensors'` check, but still appears to a user to be a safetensors
file.
For example, this might be something like `.Safetensors`,
`.SAFETENSORS`, `SafeTensors`, etc.
InvokeAI will happily import the file in the Model Manager and execute
the payload.
# Proof of Concept
1. Create a malicious pickle file.
(https://gist.github.com/CodeZombie/27baa20710d976f45fb93928cbcfe368)
2. Rename the `.ckpt` extension to some variation of `.Safetensors`,
ensuring there is a capital letter anywhere in the extension (eg.
`malicious_pickle.SAFETENSORS`)
3. Import the 'model' like you would normally with any other safetensors
file with the Model Manager.
4. Upon trying to select the model in the web ui, it will be loaded (or
attempt to be converted to a Diffuser) with `torch.load` and the payload
will execute.
# The Fix
This pull request changes the logic InvokeAI uses to decide which model
loader to use so that the safe behavior is the default. Instead of
loading as a pickle if the extension is not exactly `.safetensors`, it
will now **always** load as a safetensors file unless the extension is
**exactly** `.ckpt`.
# Notes:
I think support for pickle files should be totally dropped ASAP as a
matter of security, but I understand that there are reasons this would
be difficult.
In the meantime, I think `RestrictedUnpickler` or something similar
should be implemented as a replacement for `torch.load`, as this
significantly reduces the amount of Python methods that an attacker has
to work with when crafting malicious payloads
inside a pickle file.
Automatic1111 already uses this with some success.
(https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/master/modules/safe.py)
