Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

fix(example): minor flamingo update #33

Merged
merged 1 commit into from
Sep 13, 2023
Merged

fix(example): minor flamingo update #33

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion advanced/flamingo/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,9 @@ lep photon run \
--resource-shape gpu.a10
```

And visit [dashboard.lepton.ai](https://dashboard.lepton.ai/) to try out the model.
Optionally, add e.g. `--env OPEN_FLAMINGO_MODEL:openflamingo/OpenFlamingo-4B-vitl-rpj3b` to specify the model you would like to run. The supported model names can be found in the open-flamingo repository's README file.

You can visit [dashboard.lepton.ai](https://dashboard.lepton.ai/) to try out the model.

Note: in default, the server is protected via a token, so you won't be able to access the gradio UI. This is by design to provide adequate security. If you want to make the UI public, you can either add the `--public` argument to `lep photon run`, or update the deployment with:

Expand Down
76 changes: 62 additions & 14 deletions advanced/flamingo/photon.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,16 +1,59 @@
import base64
from io import BytesIO
import os

from typing import List, Union

from leptonai.photon import Photon, FileParam, HTTPException


# Pretrained models are obtained from https://github.com/mlfoundations/open_flamingo
# and transcribed to the following dictionary.
pretrained_models = {
"openflamingo/OpenFlamingo-3B-vitl-mpt1b": [
"ViT-L-14",
"openai",
"mosaicml/mpt-1b-redpajama-200b",
"mosaicml/mpt-1b-redpajama-200b",
1,
],
"OpenFlamingo-3B-vitl-mpt1b-langinstruct": [
"ViT-L-14",
"openai",
"mosaicml/mpt-1b-redpajama-200b-dolly",
"mosaicml/mpt-1b-redpajama-200b-dolly",
1,
],
"openflamingo/OpenFlamingo-4B-vitl-rpj3b": [
"ViT-L-14",
"openai",
"togethercomputer/RedPajama-INCITE-Base-3B-v1",
"togethercomputer/RedPajama-INCITE-Base-3B-v1",
2,
],
"openflamingo/OpenFlamingo-4B-vitl-rpj3b-langinstruct": [
"ViT-L-14",
"openai",
"togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
"togethercomputer/RedPajama-INCITE-Instruct-3B-v1",
2,
],
"openflamingo/OpenFlamingo-9B-vitl-mpt7b": [
"ViT-L-14",
"openai",
"mosaicml/mpt-7b",
"mosaicml/mpt-7b",
4,
],
}


class Flamingo(Photon):
requirement_depdency = ["open-flamingo", "huggingface-hub", "Pillow", "requests"]
requirement_dependency = ["open-flamingo", "huggingface-hub", "Pillow", "requests"]

IMAGE_TOKEN = "<image>"
END_OF_TEXT_TOKEN = "<|endofchunk|>"
DEFAULT_MODEL = "openflamingo/OpenFlamingo-3B-vitl-mpt1b"

def init(self):
from open_flamingo import create_model_and_transforms
Expand All @@ -22,25 +65,30 @@ def init(self):
else:
self.device = "cpu"

model, image_processor, tokenizer = create_model_and_transforms(
clip_vision_encoder_path="ViT-L-14",
clip_vision_encoder_pretrained="openai",
lang_encoder_path="anas-awadalla/mpt-1b-redpajama-200b",
tokenizer_path="anas-awadalla/mpt-1b-redpajama-200b",
cross_attn_every_n_layers=1,
model_name = os.environ.get("OPEN_FLAMINGO_MODEL", self.DEFAULT_MODEL)
try:
model_spec = pretrained_models[model_name]
except KeyError:
raise KeyError(
f"Model {model_name} not found in pretrained_models. Available models:"
f" {pretrained_models.keys()}"
)

self.model, self.image_processor, self.tokenizer = create_model_and_transforms(
clip_vision_encoder_path=model_spec[0],
clip_vision_encoder_pretrained=model_spec[1],
lang_encoder_path=model_spec[2],
tokenizer_path=model_spec[3],
cross_attn_every_n_layers=model_spec[4],
)

checkpoint_path = hf_hub_download(
"openflamingo/OpenFlamingo-3B-vitl-mpt1b", "checkpoint.pt"
)
model.load_state_dict(torch.load(checkpoint_path), strict=False)
model = model.to(self.device)

tokenizer.padding_side = "left"
self.model.load_state_dict(torch.load(checkpoint_path), strict=False)
self.model = self.model.to(self.device)

self.model = model
self.image_processor = image_processor
self.tokenizer = tokenizer
self.tokenizer.padding_side = "left"

def _img_param_to_img(self, param):
from PIL import Image
Expand Down