examples: support LLaVA v1.5 (multimodal model) (#3436)

* WIP: start implementing LLaVA

* rm scratch buf for now, will revert after cleanup

* LLaVA image encoder is working. will combine with llama

* Add llava inference code, but it's buggy. debugging

* LLaVA is working e2e, needs to optimize memory allocation + cleanup

* Use ggml_allocr + rm unnecessary code

* fix: crlf -> lf

* fix: new line at EoF

* fix: trailing whitespace

* Add readme

* Update readme

* Some cleanup

* Are you happy editorconfig?

* rm unused batch image preprocessing

* rm unused import

* fix: rm designated initializers

* introduce pad-to-square mode for non-square images

* are you happy editorconfig?

* gitignore /llava

* Handle cases where image file does not exist

* add llava target to Makefile

* add support for 13b model variant

* Maybe seed is unlucky?

* Check if apples are compared to apples

* are you happy editorconfig?

* Use temperature = 0.1 by default

* command line: use gpt_params_parse()

* minor

* handle default n_predict

* fix typo

* llava : code formatting, rename files, fix compile warnings

* do not use Wno-cast-qual for MSVC

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

1 files changed, 30 insertions, 0 deletions

diff --git a/examples/llava/llava-surgery.py b/examples/llava/llava-surgery.py
new file mode 100644
index 00000000..26294d9b
--- /dev/null
+++ b/examples/llava/llava-surgery.py
@@ -0,0 +1,30 @@
+import argparse
+import glob
+import os
+import torch
+
+
+ap = argparse.ArgumentParser()
+ap.add_argument("-m", "--model", help="Path to LLaVA v1.5 model")
+args = ap.parse_args()
+
+# find the model part that includes the the multimodal projector weights
+path = sorted(glob.glob(f"{args.model}/pytorch_model*.bin"))[-1]
+checkpoint = torch.load(path)
+
+# get a list of mm tensor names
+mm_tensors = [k for k, v in checkpoint.items() if k.startswith("model.mm_projector")]
+
+# store these tensors in a new dictionary and torch.save them
+projector = {name: checkpoint[name] for name in mm_tensors}
+torch.save(projector, f"{args.model}/llava.projector")
+
+# remove these tensors from the checkpoint and save it again
+for name in mm_tensors:
+    del checkpoint[name]
+
+torch.save(checkpoint, path)
+
+print("Done!")
+print(f"Now you can convert {args.model} to a a regular LLaMA GGUF file.")
+print(f"Also, use {args.model}/llava.projector to prepare a llava-encoder.gguf file.")