llm001 L05 sft: 1 epoch

README.md

@@ -0,0 +1,45 @@
+---
+base_model: complexly/olmo3-190m-zh-sft
+license: apache-2.0
+language:
+- zh
+tags:
+- llm001
+- olmo3
+- chinese
+- sft
+- supervised-finetuning
+---
+
+# complexly/olmo3-190m-zh-sft
+
+SFT（有监督微调）版本：基于complexly/olmo3-190m-zh-continue，
+使用对话格式数据进行微调，学习指令遵循能力。
+
+## 数据来源
+- 训练数据：cmz1024/llm101-olmo3-zh-demo-data
+- 子路径：sft/sft_t2t_mini.jsonl
+
+## 训练配置
+- Learning Rate：5.0e-5
+- Warmup：5%
+- Epochs：1 epoch
+- Max Seq Length：2048
+- 使用 assistant_only_loss（仅对 assistant 部分计算 loss）
+- per_device_train_batch_size: 24
+- packing: true
+
+## 用法
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers import pipeline
+
+model = AutoModelForCausalLM.from_pretrained("{target_repo}")
+tok = AutoTokenizer.from_pretrained("{target_repo}")
+
+# 使用 chat template
+messages = [{{"role": "user", "content": "你好，请介绍一下北京"}}]
+inputs = tok.apply_chat_template(messages, return_tensors="pt")
+outputs = model.generate(inputs, max_new_tokens=200)
+print(tok.decode(outputs[0]))
+```

chat_template.jinja

@@ -0,0 +1,8 @@
+{% 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' %}{% generation %}<|im_start|>assistant
+{{ message['content'] }}<|im_end|>
+{% endgeneration %}{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant
+{% endif %}

config.json

@@ -0,0 +1,44 @@
+{
+  "architectures": [
+    "Olmo3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 2,
+  "dtype": "bfloat16",
+  "eos_token_id": 0,
+  "hidden_act": "silu",
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 4096,
+  "model_type": "olmo3",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "num_key_value_heads": 12,
+  "pad_token_id": 1,
+  "rms_norm_eps": 1e-06,
+  "rope_parameters": {
+    "rope_theta": 500000,
+    "rope_type": "default"
+  },
+  "sliding_window": 4096,
+  "tie_word_embeddings": false,
+  "transformers_version": "5.8.0",
+  "use_cache": false,
+  "vocab_size": 48000
+}

generation_config.json

@@ -0,0 +1,12 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 2,
+  "eos_token_id": [
+    0
+  ],
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "pad_token_id": 1,
+  "transformers_version": "5.8.0",
+  "use_cache": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f162dc3cf093bc380f15f1ea8269a2bae56e7f46b045eb562d8732baa7355263
+size 374038864

tokenizer_config.json

@@ -0,0 +1,12 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<|bos|>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 4096,
+  "pad_token": "<|pad|>",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": "<|unk|>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32bc6ed63d160c7ed3e2c867b671df53859981927aaa07bea1a5511b33a791aa
+size 5368