llm001 L04 Nano: 26M, 20-step test

README.md

@@ -0,0 +1,38 @@
+---
+base_model: cmz1024/olmo3-190m-zh-nano
+license: apache-2.0
+language:
+- zh
+tags:
+- llm001
+- olmo3
+- chinese
+- pretrained
+---
+
+# OLMo3-190M-zh-nano
+
+为零基础 AI 大模型研发训练营（llm001）L04 Nano 模型（26M 参数，20 步测试训练）。
+
+## 模型配置（Full 的等比缩小版）
+
+| 参数 | Full (190M) | Nano (26M) |
+|------|-------------|------------|
+| hidden_size | 768 | 192 |
+| num_layers | 12 | 6 |
+| num_heads | 12 | 3 |
+| intermediate_size | 3072 | 768 |
+| vocab_size | 48000 | 48000 |
+
+## 训练配置
+
+- 数据：cmz1024/llm101-olmo3-zh-demo-data (500M tokens)
+- 训练：H100, max_steps=20, bs=32×4=128, lr=1e-3, bf16
+
+## 用法
+
+```python
+from transformers import AutoModelForCausalLM, AutoTokenizer
+model = AutoModelForCausalLM.from_pretrained("cmz1024/olmo3-190m-zh-nano")
+tok = AutoTokenizer.from_pretrained("cmz1024/olmo3-190m-zh-nano")
+```

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 +1,38 @@
+{
+  "architectures": [
+    "Olmo3ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 2,
+  "dtype": "float32",
+  "eos_token_id": 0,
+  "hidden_act": "silu",
+  "hidden_size": 192,
+  "initializer_range": 0.02,
+  "intermediate_size": 768,
+  "layer_types": [
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "max_position_embeddings": 4096,
+  "model_type": "olmo3",
+  "num_attention_heads": 3,
+  "num_hidden_layers": 6,
+  "num_key_value_heads": 3,
+  "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": "5.8.1",
+  "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": "5.8.1",
+  "use_cache": true
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0445ecf19a82ec5bcf3de799d9ebea0eb9fa37a947420cdecdc0b6005ffd9d33
+size 87910488

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 sha256:6b19cb2d9aa53d2b86a2bb70f5b499c22f9d24e23f3e3370915fcb92ab09b6aa
+size 4856