 AlpinDale
|
9e9515f39a
fix: feature size calculation for Llava-next
|
il y a 4 mois |
|
AlpinDale
|
ad68d149d8
chore: refactor and decouple phi3v image embedding
|
il y a 4 mois |
|
AlpinDale
|
e26a4ac698
chore: avoid loading the unused layers and init the VLM up to the required feature space
|
il y a 4 mois |
|
AlpinDale
|
7e99578712
fix: cleanup validation and update docs for vlm
|
il y a 4 mois |
|
AlpinDale
|
526163003d
fix: improve consistency between feature size calc and dummy data for profiling
|
il y a 4 mois |
|
AlpinDale
|
c11a8bdaad
fix: calculate max number of multi-modal tokens automatically
|
il y a 4 mois |
|
AlpinDale
|
4f7d212b70
feat: remove vision language config
|
il y a 4 mois |
|
AlpinDale
|
4599c98f99
feat: dynamic image size support for VLMs
|
il y a 4 mois |
 AlpinDale
|
0f4a9ee77b
quantized lm_head (#582)
|
il y a 4 mois |
|
AlpinDale
|
ae04f57ec1
feat: Pipeline Parallel support (#581)
|
il y a 4 mois |
|
AlpinDale
|
3a0fdf7b9b
chore: remove `image_input_type` from VLM config
|
il y a 4 mois |
|
AlpinDale
|
4cdc810b1c
fix: minor TP issues with vision models
|
il y a 5 mois |
|
AlpinDale
|
c0c336aaa3
refactor: registry for processing model inputs; quick_gelu; clip model support
|
il y a 5 mois |
|
AlpinDale
|
c5d8028668
fix: no need to redefine supports_vision and supports_lora in model class
|
il y a 5 mois |
|
AlpinDale
|
b81966c0da
fix: missed phi3v
|
il y a 5 mois |
|
AlpinDale
|
5974495461
chore: phi3v resize for dynamic shape
|
il y a 5 mois |
|
AlpinDale
|
79b1c0b861
fix: do not error our if two processes do not agree on p2p capability
|
il y a 5 mois |
|
AlpinDale
|
e6d70101b3
feat: add support for phi-3 vision model
|
il y a 5 mois |
