vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update convert-image-encoder-to-gguf.py

Browse source
This commit is contained in:
John 2024-02-02 00:25:08 +01:00 • committed by GitHub
parent 97dda1e098
commit 8ebdaec761
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 34 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

35
examples/llava/convert-image-encoder-to-gguf.py
View file

@ -203,8 +203,41 @@ if has_vision_encoder:
fout.add_float32(k(KEY_ATTENTION_LAYERNORM_EPS, VISION), v_hparams["layer_norm_eps"])
block_count = v_hparams["num_hidden_layers"] - 1 if has_llava_projector else v_hparams["num_hidden_layers"]
fout.add_uint32(k(KEY_BLOCK_COUNT, VISION), block_count)
# /**
# "image_grid_pinpoints": [
# [
# 336,
# 672
# ],
# [
# 672,
# 336
# ],
# [
# 672,
# 672
# ],
# [
# 1008,
# 336
# ],
# [
# 336,
# 1008
# ]
# ],
# Flattened:
# [
# 336, 672,
# 672, 336,
# 672, 672,
# 1008, 336,
# 336, 1008
# ]
# *
# */
if "image_grid_pinpoints" in v_hparams:
# no nested array - flatten it
# flatten it
image_grid_pinpoints = []
for pinpoint in v_hparams["image_grid_pinpoints"]:
image_grid_pinpoints.extend(pinpoint)