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

tool-call: don't use -fa w/ Mistral-Nemo (hard crashes?)

Browse source
This commit is contained in:
ochafik 2024-10-31 13:53:11 +00:00
parent f5b7825595
commit c773516d57
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
scripts/fetch_server_test_models.py
View file

@ -69,7 +69,7 @@ for m in sorted(list(models), key=lambda m: m.hf_repo):
continue continue
print(f'# Ensuring model at {m.hf_repo} / {m.hf_file} is fetched') print(f'# Ensuring model at {m.hf_repo} / {m.hf_file} is fetched')
cmd = [cli_path, '-hfr', m.hf_repo, '-hff', m.hf_file, '-n', '1', '-p', 'Hey', '--no-warmup', '--log-disable'] cmd = [cli_path, '-hfr', m.hf_repo, '-hff', m.hf_file, '-n', '1', '-p', 'Hey', '--no-warmup', '--log-disable']
if m.hf_file != 'tinyllamas/stories260K.gguf': if m.hf_file != 'tinyllamas/stories260K.gguf' and not m.hf_file.startswith('Mistral-Nemo'):
cmd.append('-fa') cmd.append('-fa')
try: try:
subprocess.check_call(cmd) subprocess.check_call(cmd)