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

Fix External CLIP Converter #8960

Merged
merged 1 commit into from
Apr 18, 2024
Merged

Fix External CLIP Converter #8960

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
16 changes: 14 additions & 2 deletions examples/multimodal/vision_language_foundation/clip/convert_external_clip_to_nemo.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,18 @@
If converting from Hugging Face, set the version to `huggingface` and the architecture (`arch`) to the Hugging Face model name (e.g., `yuvalkirstain/PickScore_v1`).

Additionally, provide a NeMo hparams file with the correct model architecture arguments. Refer to examples/multimodal/foundation/clip/conf/megatron_clip_config.yaml.

After conversion, you can verify with the following command:

wget https://upload.wikimedia.org/wikipedia/commons/0/0f/1665_Girl_with_a_Pearl_Earring.jpg
python /opt/NeMo/examples/multimodal/vision_language_foundation/clip/megatron_clip_infer.py \
model.restore_from_path=./open_clip.nemo \
image_path=./1665_Girl_with_a_Pearl_Earring.jpg \
texts='["a dog", "a boy", "a girl"]'

It should generate a high probability for "a girl" tag, e.g.
Given image's CLIP text probability: [('a dog', 0.0049710185), ('a boy', 0.002258187), ('a girl', 0.99277073)]

"""

import os
Expand Down Expand Up @@ -164,8 +176,8 @@ def mapping_hf_state_dict(hf_model):
".pre_layrnorm.bias": ".preprocess_layernorm.bias",
".post_layernorm.weight": ".transformer.final_layernorm.weight",
".post_layernorm.bias": ".transformer.final_layernorm.bias",
".backbone.embeddings.position_embedding.weight": ".backbone.position_embeddings.weight",
".language_model.embeddings.position_embedding.weight": ".language_model.embedding.position_embeddings.weight",
".backbone.embeddings.position_embedding.weight": ".backbone.position_embeddings",
".language_model.embeddings.position_embedding.weight": ".language_model.embedding.position_embeddings",
".embeddings.class_embedding": ".cls_token",
".backbone.embeddings.patch_embedding.weight": ".backbone.conv1.weight",
".final_layer_norm.weight": ".encoder.final_layernorm.weight",
Expand Down
Loading