minimize diffs

Author: ochafik

common/chat.cpp

@@ -377,8 +377,8 @@ static common_chat_params common_chat_params_init_command_r7b(const common_chat_
     return data;
 }
 static common_chat_msg common_chat_parse_command_r7b(const std::string & input) {
-    static std::regex response_regex("<\\|START_RESPONSE\\|>([\\s\\S\\n\\r]*?)<\\|END_RESPONSE\\|>");
-    static std::regex thought_action_regex("<\\|START_THINKING\\|>([\\s\\S\\n\\r]*)<\\|END_THINKING\\|><\\|START_ACTION\\|>([\\s\\S\\n\\r]*?)<\\|END_ACTION\\|>");
+    static std::regex response_regex("<\\|START_RESPONSE\\|>(.*?)<\\|END_RESPONSE\\|>");
+    static std::regex thought_action_regex("<\\|START_THINKING\\|>([\\s\\S\\n\\r]*?)<\\|END_THINKING\\|><\\|START_ACTION\\|>([\\s\\S\\n\\r]*?)<\\|END_ACTION\\|>");
     std::smatch match;
 
     common_chat_msg result;
@@ -576,7 +576,7 @@ static common_chat_params common_chat_params_init_deepseek_r1(const common_chat_
 }
 static common_chat_msg common_chat_parse_deepseek_r1(const std::string & input) {
     static std::regex trigger_regex("<｜tool▁calls▁begin｜>");
-    static std::regex function_regex(R"(<｜tool▁call▁begin｜>function<｜tool▁sep｜>([^\n]+)\n```json\n)");
+    static std::regex function_regex("<｜tool▁call▁begin｜>function<｜tool▁sep｜>([^\n]+)\n```json\n");
     static std::regex close_regex("```<｜tool▁call▁end｜>");
     static std::regex think_regex(R"(<think>([\s\S\n]*)</think>([\s\S\r\n]*))");
     auto msg = parse_json_tool_calls(input, trigger_regex, function_regex, close_regex);

common/common.cpp

@@ -1869,16 +1869,9 @@ std::string common_chat_format_example(const common_chat_template & tmpl, bool u
     return common_chat_apply_template(tmpl, msgs, true, use_jinja);
 }
 
-#define CHATML_TEMPLATE_SRC \
-    "{%- for message in messages -%}\n" \
-    "  {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>\n' -}}\n" \
-    "{%- endfor -%}\n" \
-    "{%- if add_generation_prompt -%}\n" \
-    "  {{- '<|im_start|>assistant\n' -}}\n" \
-    "{%- endif -%})"
-
 common_chat_templates common_chat_templates_from_model(const struct llama_model * model, const std::string & chat_template_override)
 {
+    auto vocab = llama_model_get_vocab(model);
     std::string default_template_src = chat_template_override == "chatml" ? CHATML_TEMPLATE_SRC : chat_template_override;
     std::string template_tool_use_src = chat_template_override == "chatml" ? CHATML_TEMPLATE_SRC : "";
     bool has_explicit_template = !chat_template_override.empty();
@@ -1908,11 +1901,6 @@ common_chat_templates common_chat_templates_from_model(const struct llama_model
             )";
         }
     }
-    std::string token_bos;
-    std::string token_eos;
-    // TODO: update logic that adds BOS and EOS tokens to the tokenized prompt, in favour of the template.
-#if 0
-    auto vocab = llama_model_get_vocab(model);
     const auto get_token = [&](llama_token token, const char * name, const char * jinja_variable_name) {
         if (token == LLAMA_TOKEN_NULL) {
             if (default_template_src.find(jinja_variable_name) != std::string::npos
@@ -1924,9 +1912,8 @@ common_chat_templates common_chat_templates_from_model(const struct llama_model
             return common_token_to_piece(vocab, token, true);
         }
     };
-    token_bos = get_token(llama_vocab_bos(vocab), "BOS", "bos_token");
-    token_eos = get_token(llama_vocab_eos(vocab), "EOS", "eos_token");
-#endif
+    auto token_bos = get_token(llama_vocab_bos(vocab), "BOS", "bos_token");
+    auto token_eos = get_token(llama_vocab_eos(vocab), "EOS", "eos_token");
     return {
         has_explicit_template,
         std::make_unique<minja::chat_template>(default_template_src, token_bos, token_eos),

examples/server/tests/unit/test_tool_call.py

@@ -252,7 +252,6 @@ def test_completion_without_tool_call_slow(template_name: str, n_predict: int, t
 @pytest.mark.slow
 @pytest.mark.parametrize("hf_repo,template_override", [
     ("bartowski/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M", None),
-    ("bartowski/c4ai-command-r7b-12-2024-GGUF:Q4_K_M",   ("CohereForAI/c4ai-command-r7b-12-2024", "tool_use")),
     ("bartowski/Meta-Llama-3.1-8B-Instruct-GGUF:Q4_K_M", None),
     ("bartowski/gemma-2-2b-it-GGUF:Q4_K_M",              None),
     ("bartowski/Phi-3.5-mini-instruct-GGUF:Q4_K_M",      None),
@@ -263,9 +262,8 @@ def test_completion_without_tool_call_slow(template_name: str, n_predict: int, t
     ("bartowski/functionary-small-v3.2-GGUF:Q8_0",       ("meetkai/functionary-medium-v3.2", None)),
     ("bartowski/Llama-3.2-3B-Instruct-GGUF:Q4_K_M",      ("meta-llama/Llama-3.2-3B-Instruct", None)),
     # ("bartowski/Llama-3.2-1B-Instruct-GGUF:Q4_K_M", ("meta-llama/Llama-3.2-3B-Instruct", None)),
-    # ("bartowski/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M", None),
 ])
-def test_weather(hf_repo: str, template_override: Tuple[str, str | None] | None):
+def test_weather_tool_call(hf_repo: str, template_override: Tuple[str, str | None] | None):
     global server
     n_predict = 512
     server.n_slots = 1
@@ -313,9 +311,8 @@ def test_weather(hf_repo: str, template_override: Tuple[str, str | None] | None)
     (None,                 "bartowski/Hermes-2-Pro-Llama-3-8B-GGUF:Q4_K_M",      ("NousResearch/Hermes-2-Pro-Llama-3-8B", "tool_use")),
     (None,                 "bartowski/Hermes-3-Llama-3.1-8B-GGUF:Q4_K_M",   ("NousResearch-Hermes-3-Llama-3.1-8B", "tool_use")),
     (None,                 "bartowski/Mistral-Nemo-Instruct-2407-GGUF:Q4_K_M", None),
-    # (None,                 "bartowski/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M", None),
 ])
-def test_hello_world(expected_arguments_override: str | None, hf_repo: str, template_override: Tuple[str, str | None] | None):
+def test_hello_world_tool_call(expected_arguments_override: str | None, hf_repo: str, template_override: Tuple[str, str | None] | None):
     global server
     server.n_slots = 1
     server.jinja = True