diff options
| author | Pedro Cuenca <pedro@huggingface.co> | 2024-04-21 13:50:41 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-04-21 14:50:41 +0300 |
| commit | b97bc3966e852adb626c90be64fd48282800f504 (patch) | |
| tree | 178656d15821205889fa03ec603c7327facbb265 /examples/server | |
| parent | b8109bc0139f15a5b321909f47510b89dca47ffc (diff) | |
llama : support Llama 3 HF conversion (#6745)
* Support Llama 3 conversion
The tokenizer is BPE.
* style
* Accept suggestion
Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com>
* llama : add llama_token_is_eog()
ggml-ci
* llama : auto-detect more EOT tokens when missing in KV data
* convert : replacing EOS token is a hack
* llama : fix codegemma EOT token + add TODOs
* llama : fix model type string for 8B model
---------
Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Diffstat (limited to 'examples/server')
| -rw-r--r-- | examples/server/server.cpp | 2 | ||||
| -rw-r--r-- | examples/server/utils.hpp | 4 |
2 files changed, 1 insertions, 5 deletions
diff --git a/examples/server/server.cpp b/examples/server/server.cpp index 634e653a..25bc2963 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -1201,7 +1201,7 @@ struct server_context { }); } - if (result.tok == llama_token_eos(model)) { + if (llama_token_is_eog(model, result.tok)) { slot.stopped_eos = true; slot.has_next_token = false; diff --git a/examples/server/utils.hpp b/examples/server/utils.hpp index a8d43ac6..1a221250 100644 --- a/examples/server/utils.hpp +++ b/examples/server/utils.hpp @@ -381,10 +381,6 @@ static json oaicompat_completion_params_parse( } else { llama_params["stop"] = json_value(body, "stop", json::array()); } - // Some chat templates don't use EOS token to stop generation - // We must add their end sequences to list of stop words - llama_params["stop"].push_back("<|im_end|>"); // chatml - llama_params["stop"].push_back("<end_of_turn>"); // gemma // Handle "response_format" field if (body.contains("response_format")) { |