upload source-screen 71875 merged model

LICENSE

@@ -0,0 +1,16 @@
+Copyright (c) 2025 SK Telecom Co., Ltd. All rights reserved.
+
+Unless otherwise stated, all files in this repository (including modified model
+weights and tokenizer files) are distributed under the terms of the Apache
+License, Version 2.0 (the "License"). You may obtain a copy of the License at:
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed
+under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+
+"SK Telecom" and associated logos are trademarks of SK Telecom Co., Ltd. This
+License does not grant permission to use these trademarks without prior written
+consent.

README.md

@@ -0,0 +1,106 @@
+---
+license: apache-2.0
+license_link: https://www.apache.org/licenses/LICENSE-2.0
+base_model: skt/A.X-3.1-Light
+language:
+- ko
+library_name: transformers
+pipeline_tag: text-generation
+tags:
+- korean
+- text-generation
+- instruction-tuning
+- supervised-fine-tuning
+- merged-model
+- source-screening
+- medical-qa
+---
+
+# A.X-3.1-Light SFT Source Screen 71875 (Essential Medical 3K)
+
+이 모델은 `skt/A.X-3.1-Light`를 기반으로 AI Hub 71875 필수의료 의학지식
+데이터만 사용해 한국어 질의응답을 LoRA 방식으로 1 epoch 지도학습한
+모델입니다. 학습이 끝난 뒤 LoRA adapter를 base model에 병합한 BF16
+standalone 전체 가중치 모델이므로 추론 시 별도의 adapter가 필요하지
+않습니다. 연구 및 통제된 평가용 모델이며, 의료 관련 답변을 포함한 모든
+생성 결과는 오류가 있을 수 있어 전문적인 판단을 대체할 수 없습니다.
+
+## Model details
+
+- Model name: `A.X-3.1-Light SFT Source Screen 71875 (Essential Medical 3K)`
+- Base model: [`skt/A.X-3.1-Light`](https://huggingface.co/skt/A.X-3.1-Light)
+- Base model revision: `9b41bb2406472634d8812c0b8931fa40fa9a6c3a`
+- Fine-tuning: LoRA supervised fine-tuning, merged into base weights
+- Weight format: BF16 `safetensors`
+- Architecture: unchanged Llama causal language model architecture
+- Chat template: bundled A.X tokenizer chat template
+- Custom model code: none; standard Transformers/vLLM loading is intended
+
+## Training data
+
+Training used only AI Hub dataset **71875, 필수의료 의학지식 데이터**. The
+training split contains 3,000 selected examples and the separate development
+split contains 300 examples. No v0.21 mixture, other AI Hub dataset, public
+benchmark question, benchmark answer, or evaluation artifact was used as SFT
+data or included in this repository.
+
+| Source | Training examples |
+|---|---:|
+| Category 14 | 895 |
+| Category 15 | 895 |
+| Category 16 | 315 |
+| Category 17 | 895 |
+| **Total** | **3,000** |
+
+The output contract is answer-first (`정답: ...`). Depending on the source
+question, the target is a label, number, short answer, or concise explanation.
+Examples over 2,048 chat-template tokens were excluded rather than truncated;
+the training summary reports zero runtime truncation. The applicable AI Hub
+terms of use remain in force. [AI Hub dataset 71875](https://www.aihub.or.kr/aihubdata/data/view.do?currMenu=115&topMenu=100&aihubDataSe=realm&dataSetSn=71875)
+
+## Training configuration
+
+- Epochs: 1
+- Optimizer steps: 375
+- Maximum sequence length: 2,048
+- Precision: BF16
+- Per-device batch size: 1
+- Gradient accumulation: 8 (effective batch size 8)
+- Learning rate: `5e-5`
+- Scheduler: cosine; warmup ratio `0.03` (11 steps)
+- Weight decay: `0.01`
+- Random seed: 42
+- LoRA rank / alpha / dropout: 16 / 32 / 0.05
+- LoRA target modules: `q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
+- Objective: assistant-token causal language-model cross entropy
+- Mean target length: 10.23 tokens; median: 5 tokens
+- Total supervised target tokens: 30,703
+- Final training loss: `0.5942968483`
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model_id = "youngseok12/AX-3.1-Light-sft_source_screen_71875_3000"
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    torch_dtype="auto",
+    device_map="auto",
+)
+```
+
+Use the bundled tokenizer chat template for conversational inference. The
+repository is a merged full model and does not require PEFT adapter loading.
+
+## Limitations and license
+
+This model is derived from the Apache-2.0 licensed `skt/A.X-3.1-Light` model;
+the base model notices and SK Telecom trademark terms also apply. AI Hub terms
+apply to the source dataset. See [`LICENSE`](LICENSE) and the base model
+repository for the applicable terms.
+
+The model can produce incorrect, incomplete, biased, or poorly formatted
+answers. It has not been validated as a medical device or professional medical
+advice system and must not be used as the sole basis for clinical decisions.

chat_template.jinja

@@ -0,0 +1,72 @@
+{%- if tools is iterable and tools | length > 0 %}
+                    {{- '<|im_start|><|system|>'}}
+                    {{- '당신은 도구 호출 기능을 갖춘 유용한 도우미입니다. 사용자의 요청을 처리하기 위해서 필요한 도구가 주어진 목록에 있는 경우 도구 호출로 응답하세요.
+필요한 도구가 목록에 없는 경우에는 도구 호출 없이 사용자가 요구한 정보를 제공하세요.
+필요한 도구가 목록에 있지만 해당 도구를 호출하는데 필요한 argument 정보가 부족한 경우 해당 정보를 사용자에게 요청하세요.
+사용자의 요청을 처리하기 위해 여러번 도구를 호출할 수 있어야 합니다.
+도구 호출 이후 도구 실행 결과를 입력으로 받으면 해당 결과를 활용하여 답변을 생성하세요.
+
+다음은 접근할 수 있는 도구들의 목록 입니다:
+<tools>
+'}}
+                    {%- for t in tools %}
+                        {{- t | tojson }}
+                        {{- '
+' }}
+                    {%- endfor %}
+                    {{- '</tools>' }}
+                    {{- '
+
+도구를 호출하려면 아래의 JSON으로 응답하세요.
+도구 호출 형식: <tool_call>{"name": 도구 이름, "arguments": dictionary 형태의 도구 인자값}</tool_call>' }}
+                    {{- '<|im_end|>' }}
+                {%- endif %}
+        
+                {%- for message in messages %}
+                    {%- if message.role == 'system' %}
+                        {{- '<|im_start|><|system|>' + message.content + '<|im_end|>'}}
+                    {%- elif message.role == 'user' %}
+                        {{- '<|im_start|><|user|>' + message.content + '<|im_end|>'}}
+                    {%- elif message.role == 'assistant' %}
+                        {{- '<|im_start|><|assistant|>'}}
+                        {%- set content = '' %}
+                        {%- if message.content is defined %}
+                            {%- set content = message.content %}
+                        {%- endif %}
+                        
+                {%- if add_generation_prompt and not (message.reasoning_content is defined and message.reasoning_content is not none) %}
+                    {%- if '</think>' in message.content %}
+                        {%- set content = message.content.split('</think>'.strip())[-1].lstrip('\n') %}
+                    {%- endif %}
+                {%- endif %}
+                
+                        {{- content}}
+                        {%- if message.tool_calls is defined %}
+                            {%- 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 }}
+                                {{- '"' }}
+                                {%- if tool_call.arguments is defined %}
+                                    {{- ', ' }}
+                                    {{- '"arguments": ' }}
+                                    {{- tool_call.arguments|tojson }}
+                                {%- endif %}
+                                {{- '}' }}
+                                {{- '</tool_call>' }}
+                            {%- endfor %}
+                        {%- endif %}
+                        {{- '<|im_end|>'}}
+        
+                    {%- elif message.role == 'tool' %}
+                        {{- '<|im_start|><|extra_id_13|><tool_output>' + message.content + '</tool_output><|im_end|>'}}
+                    {%- endif %}
+                {%- endfor %}
+                
+                    {%- if add_generation_prompt %}
+                        {{- '<|im_start|><|assistant|>' }}
+                    {%- endif %}

config.json

@@ -0,0 +1,32 @@
+{
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.1,
+  "bos_token_id": 0,
+  "dtype": "bfloat16",
+  "eos_token_id": 0,
+  "head_dim": 128,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 10880,
+  "max_position_embeddings": 32768,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": null,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_parameters": {
+    "rope_theta": 500000,
+    "rope_type": "default"
+  },
+  "tie_word_embeddings": false,
+  "transformers_version": "5.15.0",
+  "use_cache": false,
+  "vocab_size": 102400
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "bos_token_id": 0,
+  "eos_token_id": 27,
+  "max_new_tokens": 32768,
+  "pad_token_id": 1,
+  "transformers_version": "5.15.0"
+}

kds_merge_info.json

@@ -0,0 +1,5 @@
+{
+  "base_model": "AX-3.1-Light",
+  "base_model_path": "/home/youngseok3/.cache/huggingface/hub/models--skt--A.X-3.1-Light/snapshots/9b41bb2406472634d8812c0b8931fa40fa9a6c3a",
+  "adapter_path": "/home/youngseok3/KDS/runs/source_screen_71875_20260831/training_output/final_adapter"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8f50bde95b712dec0cd8b9a1302f9a68b99f5af30843e311d3aaae092e306c6
+size 14529635608

tokenizer_config.json

@@ -0,0 +1,23 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<|cls|>",
+  "eod_token": "<|endoftext|>",
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "is_local": true,
+  "local_files_only": true,
+  "mask_token": "<|mask|>",
+  "max_length": 7680,
+  "model_max_length": 32768,
+  "model_specific_special_tokens": {
+    "eod_token": "<|endoftext|>"
+  },
+  "pad_token": "<|pad|>",
+  "sep_token": "<|sep|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|unk|>",
+  "vocab_size": 102400
+}