(CUDA-only) Efficient inference using llama-mtmd-cli for high resolution images with reduced GPU VRAM usage (#17801) #17802
+660
−43
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…es while having lower GPU VRAM requirements. Implemented 3 optis to enable this: i) offload vision model weights(only) to CPU and stream to device at runtime ii) reordering LLM model init so that the CLIP model is done with encoding the image and has freed-up the VRAM memory iii) tiled flash attention to avoid 2GB/INT_MAX limit ggml_cuda_cpy for larger images
ngxson
changed the title
…t_llm_init_callback and mtmd_set_llm_context remain exported as MTMD_API as they are needed in mtmd-cli.cpp, rest other JIT functions are now internal
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Efficient VLM inference using llama-mtmd-cli for high resolution images while having lower GPU VRAM requirements. Implemented 3 optis to enable this: i) offload vision model weights(only) to CPU and stream to device at runtime ii) reordering LLM model init so that the CLIP model is done with encoding the image and has freed-up the VRAM memory iii) tiled flash attention to avoid 2GB/INT_MAX limit ggml_cuda_cpy for larger images