"values" parameter in validator receives ambiguous types

[ghost]

Checks

• I added a descriptive title to this issue
• I have searched (google, github) for similar issues and couldn't find anything
• I have read and followed the docs and still think this is a bug

Bug

Output of python -c "import pydantic.utils; print(pydantic.utils.version_info())":

python -c "import pydantic.utils; print(pydantic.utils.version_info())"                                          
             pydantic version: 1.6.1
            pydantic compiled: True
                 install path: /homes/<redacted>/.conda/envs/<redacted>/lib/python3.7/site-packages/pydantic
               python version: 3.7.8 | packaged by conda-forge | (default, Jul 31 2020, 02:25:08)  [GCC 7.5.0]
                     platform: Linux-4.9.0-0.bpo.6-amd64-x86_64-with-debian-8.11
     optional deps. installed: ['typing-extensions']

Hi,
In a validator method, when adding the values parameter, I expect it to be a map from field names to validated types.
When specifying validate_assignment = True in model config, values gets a different type (just on assignment).
This is illustrated by the example below:

import pydantic

class ModelOne(pydantic.BaseModel):
    a: int

class ModelTwo(pydantic.BaseModel):
    m: ModelOne
    b: int

    @pydantic.validator('b')
    def validate_b(cls, b, values):
        print(values)
        if 'm' in values:
            return b + values['m'].a  # this fails with AttributeError if values['m'] is a dict
        else:
            return b

    class Config:
        validate_assignment = True

model = ModelTwo(m=ModelOne(a=1), b=2)
#> {'m': ModelOne(a=1)}
model.b = 3
#> {'m': {'a': 1}}

As far as I can tell, this behavior is not documented, and I'm pretty sure it's not intended.

edit: created a PR