TymofiiNasobko/Lapa-function-calling

README.md

@@ -4,8 +4,8 @@ library_name: transformers
 model_name: Lapa-function-calling
 tags:
 - generated_from_trainer
-- sft
 - trl
+- sft
 licence: license
 ---
 
@@ -34,10 +34,10 @@ This model was trained with SFT.
 
 ### Framework versions
 
-- TRL: 0.25.1
+- TRL: 0.26.2
 - Transformers: 4.57.3
 - Pytorch: 2.6.0+cu118
-- Datasets: 4.4.1
+- Datasets: 4.4.2
 - Tokenizers: 0.22.1
 
 ## Citations

adapter_config.json

@@ -16,7 +16,7 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 512,
+  "lora_alpha": 32,
   "lora_bias": false,
   "lora_dropout": 0.05,
   "megatron_config": null,
@@ -28,20 +28,20 @@
   "peft_type": "LORA",
   "peft_version": "0.18.0",
   "qalora_group_size": 16,
-  "r": 256,
+  "r": 16,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "o_proj",
-    "v_proj",
-    "down_proj",
-    "fc2",
-    "q_proj",
+    "gate_proj",
+    "up_proj",
     "out_proj",
+    "fc2",
+    "k_proj",
+    "down_proj",
     "fc1",
-    "up_proj",
-    "gate_proj",
-    "k_proj"
+    "v_proj",
+    "o_proj",
+    "q_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d5bd45b8610d7433d4628cc89c58265703031a5b658a642c73bf3bb7de3c0ffd
-size 8774281960
+oid sha256:2a94bd8ade22dbd4bc7fdfe61013744587d42bd243f5aec473266e2205c404b8
+size 4324333160

chat_template.jinja

@@ -1,15 +1,45 @@
 {{ bos_token }}
-{%- if messages[0]['role'] == 'system' -%}
+{%- if tools -%}
+{%- set loop_messages = messages[1:] -%}
+	{{- '<start_of_turn>' + 'system' + '
+' -}}
+	{%- if messages[0]['role'] == "system" -%}
+        {%- if messages[0]['content'] is string -%}
+		    {{- messages[0]['content'] + '
+' + '
+' -}}
+        {%- else -%}
+            {{- messages[0]['content'][0]['text'] + '
+' + '
+' -}}
+        {%- endif -%}
+    {%- endif -%}
+	{{- "# Tools
+
+You may call one or more functions to assist with the user query.
+
+You are provided with function signatures within <tools></tools> XML tags:
+<tools>" -}}
+	{%- for tool in tools -%}
+		{{- "
+" -}}
+		{{- tool | tojson -}}
+	{%- endfor -%}
+    {"name": "{{ tool_call['name'] }}", "arguments": {{ tool_call['arguments'] | tojson }}}
+    {{- '<end_of_turn>' + '
+' -}}
+{%- elif messages[0]['role'] == "system" -%}
+{%- set loop_messages = messages[1:] -%}
+	{{- '<start_of_turn>' + 'system' + '
+' -}}
     {%- if messages[0]['content'] is string -%}
-        {%- set first_user_prefix = messages[0]['content'] + '
-' -%}
+        {{- messages[0]['content'] -}}
     {%- else -%}
-        {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
-' -%}
+        {{- messages[0]['content'][0]['text'] -}}
     {%- endif -%}
-    {%- set loop_messages = messages[1:] -%}
+    {{- '<end_of_turn>
+' }}
 {%- else -%}
-    {%- set first_user_prefix = "" -%}
     {%- set loop_messages = messages -%}
 {%- endif -%}
 {%- for message in loop_messages -%}
@@ -18,12 +48,25 @@
     {%- else -%}
         {%- set role = message['role'] -%}
     {%- endif -%}
-
     {{ '<start_of_turn>' + role + '
-' + (first_user_prefix if loop.first else "") }}
+' }}
     {%- if message['content'] is string -%}
         {%- if role == "model" -%}
-            {% generation %}{{ message['content'] | trim }}{% endgeneration %}
+{% generation %}
+{{- message['content'] | trim -}}
+{% if message['tool_calls'] %}
+    {% for tool_call in message['tool_calls'] %}
+        {% if tool_call['function'] is defined %}
+            {% set tool_call = tool_call['function'] %}
+        {% endif %}
+{{- "
+" -}}
+<tool_call>
+{"name": "{{ tool_call['name'] }}", "arguments": {{ tool_call['arguments'] | tojson }}}
+</tool_call>
+    {%- endfor -%}
+{%- endif -%}
+{% endgeneration %}
         {%- else -%}
             {{ message['content'] | trim }}
         {%- endif -%}
@@ -33,7 +76,21 @@
                 {{ '<start_of_image>' }}
             {%- elif item['type'] == 'text' -%}
                 {%- if role == "model" -%}
-                    {% generation %}{{ item['text'] | trim }}{% endgeneration %}
+{% generation %}
+{{- item['text'] | trim -}}
+{% if message['tool_calls'] %}
+    {% for tool_call in message['tool_calls'] %}
+        {% if tool_call['function'] is defined %}
+            {% set tool_call = tool_call['function'] %}
+        {% endif %}
+{{- "
+" -}}
+<tool_call>
+{"name": "{{ tool_call['name'] }}", "arguments": {{ tool_call['arguments'] | tojson }}}
+</tool_call>
+    {%- endfor -%}
+{%- endif -%}
+{% endgeneration %}
                 {%- else -%}
                     {{ item['text'] | trim }}
                 {%- endif -%}
@@ -51,7 +108,6 @@
 ' }}
 {%- endif -%}
 {%- endfor -%}
-
 {%- if add_generation_prompt -%}
     {{'<start_of_turn>model
 '}}

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:30e1c54445c5beccd239bb6cc38804b3a0439fb2dae5c02ee756d2bb97e018e9
-size 5816
+oid sha256:12352aff4b6f68c7f229e2cedca2ed31cbfe6e2546540c2c02f8aff4cdc9e90d
+size 5880