mirror of
https://github.com/ggml-org/llama.cpp.git
synced 2025-11-04 09:32:00 +00:00
40a751ea9a
* model-conversion : remove hardcoded /bin/bash shebangs [no ci] This commit updates the bash scripts to use env instead of using hardcoded /bin/bash in the shebang line. The motivation for this is that some systems may have bash installed in a different location, and using /usr/bin/env bash ensures that the script will use the first bash interpreter found in the user's PATH, making the scripts more portable across different environments. * model-conversion : rename script to .py [no ci] This commit renames run-casual-gen-embeddings-org.sh to run-casual-gen-embeddings-org.py to reflect its Python nature.
44 lines
1.1 KiB
Bash
Executable File
44 lines
1.1 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
|
|
set -e
|
|
|
|
MODEL_PATH="${1:-"$MODEL_PATH"}"
|
|
MODEL_NAME="${2:-$(basename "$MODEL_PATH")}"
|
|
|
|
if [ -t 0 ]; then
|
|
CPP_EMBEDDINGS="data/llamacpp-${MODEL_NAME}-embeddings.bin"
|
|
else
|
|
# Process piped JSON data and convert to binary (matching logits.cpp format)
|
|
TEMP_FILE=$(mktemp /tmp/tmp.XXXXXX.binn)
|
|
python3 -c "
|
|
import json
|
|
import sys
|
|
import struct
|
|
|
|
data = json.load(sys.stdin)
|
|
|
|
# Flatten all embeddings completely
|
|
flattened = []
|
|
for item in data:
|
|
embedding = item['embedding']
|
|
for token_embedding in embedding:
|
|
flattened.extend(token_embedding)
|
|
|
|
print(f'Total embedding values: {len(flattened)}', file=sys.stderr)
|
|
|
|
# Write as binary floats - matches logitc.cpp fwrite format
|
|
with open('$TEMP_FILE', 'wb') as f:
|
|
for value in flattened:
|
|
f.write(struct.pack('f', value))
|
|
"
|
|
CPP_EMBEDDINGS="$TEMP_FILE"
|
|
trap "rm -f $TEMP_FILE" EXIT
|
|
fi
|
|
|
|
python scripts/utils/semantic_check.py --model-path $MODEL_PATH \
|
|
--python-embeddings data/pytorch-${MODEL_NAME}-embeddings.bin \
|
|
--cpp-embeddings $CPP_EMBEDDINGS \
|
|
--prompt "Hello world today" \
|
|
--causal
|
|
|