Added --chat-template-file to llama-run #11922
Conversation
|
@ericcurtin PTAL |
engelmi
force-pushed
the
added-chat-template-file-to-llama-run
branch
from
36a0c4f to
2ec2107
Compare
engelmi
force-pushed
the
added-chat-template-file-to-llama-run
branch
from
2ec2107 to
6e50443
Compare
|
This code LGTM waiting on builds |
Should pass now. Didn't review the latest changes from master thoroughly. Not sure why I get |
examples/run/run.cpp
Outdated
| return ""; | ||
| } | ||
|
|
||
| FILE* file = ggml_fopen(chat_template_file.c_str(), "r"); |
There was a problem hiding this comment.
This would be better if it used the File class from this code. It would automatically fclose the file when necessary. I would also change fopen -> ggml_fopen in that File class.
There was a problem hiding this comment.
The file class from this code is behind the LLAMA_USE_CURL macro. So I'd need to move it since the --chat-template should also work without that macro. What do you think?
There was a problem hiding this comment.
Yeah I'd move it out of the LLAMA_USE_CURL macro. It doesn't actually do any curl stuff.
engelmi
force-pushed
the
added-chat-template-file-to-llama-run
branch
from
6e50443 to
f01a139
Compare
examples/run/run.cpp
Outdated
| printe("Error reading chat template file '%s': %s", chat_template_file.c_str(), strerror(errno)); | ||
| return ""; | ||
| } | ||
| return std::string(data.begin(), data.end()); |
There was a problem hiding this comment.
One could actually fread directly to a std::string and save a copy. It's a optimisation, a small one. A std::string is basically a vector of chars.
Never seen that before 🤷 |
Relates to: ggml-org#11178 Added --chat-template-file CLI option to llama-run. If specified, the file will be read and the content passed for overwriting the chat template of the model to common_chat_templates_from_model. Signed-off-by: Michael Engel <[email protected]>
engelmi
force-pushed
the
added-chat-template-file-to-llama-run
branch
from
f01a139 to
86f68a8
Compare
Hmm... closing and reopening it. Hopefully this fixes that error. |
2 participants
Relates to: #11178
Added --chat-template-file CLI option to llama-run. If specified, the file will be read and the content passed for overwriting the chat template of the model to common_chat_templates_from_model.
This also enables running the
granite-codemodel from ollama:Make sure to read the contributing guidelines before submitting a PR