Text Generation
Transformers
Safetensors
English
Korean
hrm_text
terminal
sft
vllm
tb2-lite
evaluation-pending
Instructions to use LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="LLM-OS-Models/KoHRM-Text-1.4B-Epoch2")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("LLM-OS-Models/KoHRM-Text-1.4B-Epoch2") model = AutoModelForCausalLM.from_pretrained("LLM-OS-Models/KoHRM-Text-1.4B-Epoch2", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/LLM-OS-Models/KoHRM-Text-1.4B-Epoch2
- SGLang
How to use LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 with Docker Model Runner:
docker model run hf.co/LLM-OS-Models/KoHRM-Text-1.4B-Epoch2
Update model card with pending TB2-lite evaluation status
Browse files
README.md
CHANGED
|
@@ -1,702 +1,167 @@
|
|
| 1 |
---
|
| 2 |
-
license: other
|
| 3 |
language:
|
| 4 |
-
- ko
|
| 5 |
- en
|
| 6 |
-
tags:
|
| 7 |
-
- kohrm
|
| 8 |
-
- epoch2-final
|
| 9 |
-
- hrm-text
|
| 10 |
-
- korean
|
| 11 |
-
- terminal
|
| 12 |
-
- tool-use
|
| 13 |
-
library_name: transformers
|
| 14 |
-
---
|
| 15 |
-
|
| 16 |
-
# KoHRM-Text-1.4B Epoch 2 Final
|
| 17 |
-
|
| 18 |
-
This repository/revision is the fixed **Epoch 2 Final** export of KoHRM-Text-1.4B.
|
| 19 |
-
|
| 20 |
-
- Epoch label: `epoch2-final`
|
| 21 |
-
- Final stage: `stage4b-korean-tool-finance-repeat`
|
| 22 |
-
- Global step: `470077`
|
| 23 |
-
- Export: EMA weights converted to `model.safetensors`
|
| 24 |
-
- Rolling latest repo: `https://huggingface.co/LLM-OS-Models/KoHRM-Text-1.4B`
|
| 25 |
-
- Raw resume checkpoints: `https://huggingface.co/LLM-OS-Models/KoHRM-Text-1.4B-raw-checkpoints`
|
| 26 |
-
|
| 27 |
-
ํ๊ตญ์ด: ์ด ์ ์ฅ์/๋ฆฌ๋น์ ์ KoHRM-Text-1.4B์ **์ํญ 2 ์๋ฃ๋ณธ**์ ๊ณ ์ ์ ์ฅํ ๊ฒ์
๋๋ค. ์ดํ ํ์ต์ด ๊ณ์ ์งํ๋์ด rolling latest๊ฐ ๋ฐ๋์ด๋ ์ด artifact๋ epoch 2 ์๋ฃ ์ง์ ์ ๊ฐ๋ฆฌํต๋๋ค.
|
| 28 |
-
|
| 29 |
-
---
|
| 30 |
-
---
|
| 31 |
-
license: apache-2.0
|
| 32 |
-
language:
|
| 33 |
- ko
|
| 34 |
-
|
|
|
|
| 35 |
tags:
|
| 36 |
-
- hrm-text
|
| 37 |
-
- korean
|
| 38 |
- terminal
|
| 39 |
-
-
|
| 40 |
-
-
|
| 41 |
-
-
|
| 42 |
-
-
|
| 43 |
-
|
| 44 |
-
pipeline_tag: text-generation
|
| 45 |
---
|
| 46 |
|
| 47 |
-
# KoHRM-Text-1.4B
|
| 48 |
-
|
| 49 |
-
**Language / ์ธ์ด:** [English](#english) | [ํ๊ตญ์ด](#korean)
|
| 50 |
-
|
| 51 |
-
<a id="english"></a>
|
| 52 |
-
|
| 53 |
-
## English
|
| 54 |
-
|
| 55 |
-
`KoHRM-Text-1.4B` is a scratch-pretrained Korean/English/code/terminal/tool-use model built from the `sapientinc/HRM-Text` PrefixLM training stack.
|
| 56 |
-
|
| 57 |
-
This is **not** a continued finetune of `sapientinc/HRM-Text-1B`. It uses a new Korean/terminal-oriented 131K byte-level BPE tokenizer and a new scratch training run.
|
| 58 |
-
|
| 59 |
-
### Current Status
|
| 60 |
-
|
| 61 |
-
This repository is a rolling **latest public model export**. Training is still in progress.
|
| 62 |
-
|
| 63 |
-
- Main repo: `LLM-OS-Models/KoHRM-Text-1.4B`
|
| 64 |
-
- Current public files: `model.safetensors`, `config.json`, tokenizer files, and this `README.md`
|
| 65 |
-
- Raw FSDP2 resume checkpoints: `LLM-OS-Models/KoHRM-Text-1.4B-raw-checkpoints`
|
| 66 |
-
- Prepared data: `LLM-OS-Models/KoHRM-Text-1.4B-prepared-data`
|
| 67 |
-
- Project code: https://github.com/LLM-OS-Models/KoHRM-text
|
| 68 |
-
- Upstream HRM-Text code: https://github.com/sapientinc/HRM-Text
|
| 69 |
-
- HRM-Text paper: https://arxiv.org/html/2605.20613
|
| 70 |
-
- Tokenizer repo: `LLM-OS-Models/HRM-Text-Ko-Terminal-Tokenizer-131K`
|
| 71 |
-
|
| 72 |
-
The main branch is overwritten with the newest converted EMA `safetensors` export as training checkpoints are uploaded. To test the latest public weight, download `revision="main"`.
|
| 73 |
-
|
| 74 |
-
### Training Method At A Glance
|
| 75 |
-
|
| 76 |
-
KoHRM-Text is best understood as **instruction pretraining from scratch**.
|
| 77 |
-
|
| 78 |
-
It is not ordinary raw-text causal LM pretraining, and it is not only a small SFT pass on top of an existing base model.
|
| 79 |
-
|
| 80 |
-
```text
|
| 81 |
-
raw data -> tokenizer -> V1Dataset -> PrefixLM batches
|
| 82 |
-
-> HRM H/L recurrence -> LM head -> response-only loss
|
| 83 |
-
```
|
| 84 |
-
|
| 85 |
-
The input context is handled as a PrefixLM prefix:
|
| 86 |
-
|
| 87 |
-
```text
|
| 88 |
-
instruction / prefix: bidirectional attention, no loss
|
| 89 |
-
response: causal attention, response-only CE loss
|
| 90 |
-
```
|
| 91 |
-
|
| 92 |
-
The architecture keeps the upstream HRM-Text recurrent design:
|
| 93 |
-
|
| 94 |
-
```text
|
| 95 |
-
H module: slower strategic state
|
| 96 |
-
L module: faster execution state
|
| 97 |
-
schedule: H2L3 recurrent computation
|
| 98 |
-
```
|
| 99 |
-
|
| 100 |
-
For a readable full explanation of the training method, architecture, PT/SFT distinction, staged continuation, and checkpoint naming, see the project document:
|
| 101 |
-
|
| 102 |
-
[MODEL_TRAINING_ARCHITECTURE_GUIDE_2026-05-28.md](MODEL_TRAINING_ARCHITECTURE_GUIDE_2026-05-28.md) in https://github.com/LLM-OS-Models/KoHRM-text
|
| 103 |
-
|
| 104 |
-
### Important Compatibility Note
|
| 105 |
-
|
| 106 |
-
The public repo currently contains the converted model weights and tokenizer, but it does **not yet** include a Hugging Face `trust_remote_code` modeling implementation for `HrmTextForCausalLM`.
|
| 107 |
-
|
| 108 |
-
What works today:
|
| 109 |
-
|
| 110 |
-
- Download the latest public weights.
|
| 111 |
-
- Load the tokenizer directly with `tokenizers.Tokenizer.from_file("tokenizer.json")`.
|
| 112 |
-
- Inspect `config.json`.
|
| 113 |
-
- Verify `model.safetensors` on CPU or Colab T4.
|
| 114 |
-
|
| 115 |
-
What is not supported yet in plain Transformers:
|
| 116 |
-
|
| 117 |
-
- `AutoModelForCausalLM.from_pretrained("LLM-OS-Models/KoHRM-Text-1.4B")`
|
| 118 |
-
- One-line hosted text generation from this repo
|
| 119 |
-
|
| 120 |
-
Expected reason: `model_type: "hrm_text"` is a custom HRM-Text architecture. Public generation will require adding the compatible `HrmTextForCausalLM` remote-code files to this model repo or releasing a standard wrapper.
|
| 121 |
-
|
| 122 |
-
### Model Details
|
| 123 |
-
|
| 124 |
-
| Field | Value |
|
| 125 |
-
|---|---:|
|
| 126 |
-
| Model id | `LLM-OS-Models/KoHRM-Text-1.4B` |
|
| 127 |
-
| Standard name | `KoHRM-Text-1.4B` |
|
| 128 |
-
| Training origin | scratch |
|
| 129 |
-
| Architecture family | HRM-Text PrefixLM |
|
| 130 |
-
| Architecture size | `XL` |
|
| 131 |
-
| Parameters | 1,384,120,320 |
|
| 132 |
-
| Context length | 4,096 tokens |
|
| 133 |
-
| Training dtype | bfloat16 |
|
| 134 |
-
| Public export dtype | bfloat16 EMA `safetensors` |
|
| 135 |
-
| Tokenizer | byte-level BPE, NFC normalization |
|
| 136 |
-
| Vocabulary size | 131,072 |
|
| 137 |
-
| Objective | PrefixLM response-only loss |
|
| 138 |
-
| Optimizer | Adam-atan2 from upstream HRM-Text |
|
| 139 |
-
| EMA | 0.9999 |
|
| 140 |
|
| 141 |
-
|
| 142 |
|
| 143 |
-
|
| 144 |
-
{
|
| 145 |
-
"model_type": "hrm_text",
|
| 146 |
-
"architectures": ["HrmTextForCausalLM"],
|
| 147 |
-
"vocab_size": 131072,
|
| 148 |
-
"hidden_size": 1536,
|
| 149 |
-
"num_hidden_layers": 32,
|
| 150 |
-
"num_attention_heads": 12,
|
| 151 |
-
"max_position_embeddings": 4096,
|
| 152 |
-
"prefix_lm": true
|
| 153 |
-
}
|
| 154 |
-
```
|
| 155 |
-
|
| 156 |
-
### Compared With The HRM-Text Paper
|
| 157 |
-
|
| 158 |
-
This run can take longer than the paper recipe even on 8 x H200 because the setup is not identical:
|
| 159 |
|
| 160 |
-
-
|
| 161 |
-
-
|
| 162 |
-
-
|
| 163 |
-
-
|
| 164 |
-
-
|
| 165 |
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
### Tokenizer
|
| 169 |
-
|
| 170 |
-
The tokenizer was trained for Korean, English, code, shell/terminal text, and JSON/tool-call formats. It keeps common chat/tool special tokens as stable single tokens where possible.
|
| 171 |
-
|
| 172 |
-
| Sample bucket | chars/token |
|
| 173 |
-
|---|---:|
|
| 174 |
-
| Korean general text | 2.60 |
|
| 175 |
-
| Korean legal text | 2.36 |
|
| 176 |
-
| Korean terminal instruction | 2.18 |
|
| 177 |
-
| shell command | 2.68 |
|
| 178 |
-
| tool-call JSON | 3.32 |
|
| 179 |
-
| Python code | 3.37 |
|
| 180 |
-
| English | 4.40 |
|
| 181 |
-
|
| 182 |
-
Formatting tokens:
|
| 183 |
-
|
| 184 |
-
```text
|
| 185 |
-
<|im_start|> instruction start
|
| 186 |
-
<|im_end|> instruction end
|
| 187 |
-
<|box_end|> response/end marker
|
| 188 |
-
<|object_ref_start|> direct condition
|
| 189 |
-
<|object_ref_end|> chain-of-thought style condition
|
| 190 |
-
<|quad_start|> noisy condition
|
| 191 |
-
<|quad_end|> synthetic condition
|
| 192 |
-
```
|
| 193 |
|
| 194 |
-
|
| 195 |
|
| 196 |
-
```
|
| 197 |
-
|
|
|
|
| 198 |
```
|
| 199 |
|
| 200 |
-
|
| 201 |
|
| 202 |
-
|
|
|
|
|
|
|
|
|
|
| 203 |
|
| 204 |
-
|
| 205 |
-
|
| 206 |
-
The notebook downloads the latest public files and runs a short generation test on a Colab T4.
|
| 207 |
-
|
| 208 |
-
It intentionally avoids `transformers`, `AutoTokenizer`, and `AutoModelForCausalLM`. Instead, it uses:
|
| 209 |
-
|
| 210 |
-
- `tokenizers.Tokenizer.from_file("tokenizer.json")`
|
| 211 |
-
- `safetensors.torch.load_file("model.safetensors")`
|
| 212 |
-
- `kohrm_colab_generate.py`, a small PyTorch SDPA runtime for the HRM-Text architecture
|
| 213 |
-
|
| 214 |
-
```python
|
| 215 |
-
!pip -q install -U huggingface_hub hf_transfer tokenizers safetensors
|
| 216 |
-
```
|
| 217 |
|
| 218 |
```python
|
| 219 |
-
from
|
| 220 |
-
import
|
| 221 |
-
|
| 222 |
-
|
| 223 |
-
|
| 224 |
-
|
| 225 |
-
|
| 226 |
-
|
| 227 |
-
|
| 228 |
-
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
| 232 |
-
|
| 233 |
-
|
| 234 |
-
"tokenizer.json",
|
| 235 |
-
"tokenizer_config.json",
|
| 236 |
-
"special_tokens_map.json",
|
| 237 |
-
"model.safetensors",
|
| 238 |
-
"kohrm_colab_generate.py",
|
| 239 |
-
],
|
| 240 |
-
))
|
| 241 |
-
|
| 242 |
-
print("Downloaded to:", repo_dir)
|
| 243 |
-
config = json.loads((repo_dir / "config.json").read_text())
|
| 244 |
-
print("model_type:", config["model_type"])
|
| 245 |
-
print("hidden_size:", config["hidden_size"])
|
| 246 |
-
print("vocab_size:", config["vocab_size"])
|
| 247 |
-
print("context:", config["max_position_embeddings"])
|
| 248 |
-
|
| 249 |
-
tokenizer = Tokenizer.from_file(str(repo_dir / "tokenizer.json"))
|
| 250 |
-
wrapped = "<|im_start|><|object_ref_start|>ํ๊ตญ์ด๋ก ํ์ฌ ๋๋ ํฐ๋ฆฌ์์ ๊ฐ์ฅ ํฐ ํ์ผ 10๊ฐ๋ฅผ ์ฐพ๋ ๋ช
๋ น์ ์๋ ค์ฃผ์ธ์.<|im_end|>"
|
| 251 |
-
ids = tokenizer.encode(wrapped).ids
|
| 252 |
-
print("prompt tokens:", len(ids))
|
| 253 |
-
print("first token ids:", ids[:20])
|
| 254 |
-
|
| 255 |
-
spec = importlib.util.spec_from_file_location(
|
| 256 |
-
"kohrm_colab_generate",
|
| 257 |
-
repo_dir / "kohrm_colab_generate.py",
|
| 258 |
)
|
| 259 |
-
kohrm = importlib.util.module_from_spec(spec)
|
| 260 |
-
sys.modules["kohrm_colab_generate"] = kohrm
|
| 261 |
-
spec.loader.exec_module(kohrm)
|
| 262 |
|
| 263 |
-
|
| 264 |
-
|
| 265 |
-
"
|
| 266 |
-
|
| 267 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 268 |
temperature=0.0,
|
|
|
|
|
|
|
|
|
|
| 269 |
)
|
| 270 |
-
|
| 271 |
-
|
| 272 |
-
|
| 273 |
-
Expected result:
|
| 274 |
-
|
| 275 |
-
- `model_type` should be `hrm_text`.
|
| 276 |
-
- `vocab_size` should be `131072`.
|
| 277 |
-
- The helper should load the 1.38B public `model.safetensors` export.
|
| 278 |
-
- On Colab T4, generation runs in fp16 through PyTorch scaled-dot-product attention.
|
| 279 |
-
- First generation can take a few minutes because it downloads and loads the full weight file.
|
| 280 |
-
|
| 281 |
-
Prompt format used by the helper:
|
| 282 |
-
|
| 283 |
-
```text
|
| 284 |
-
<|im_start|><|object_ref_start|>PROMPT<|im_end|>
|
| 285 |
-
```
|
| 286 |
-
|
| 287 |
-
Plain `AutoModelForCausalLM.generate()` is still not the supported path. This model is a custom `hrm_text` architecture, so ordinary Transformers generation requires a future `trust_remote_code` wrapper. Use the notebook/helper above for public `model.safetensors` generation today.
|
| 288 |
-
|
| 289 |
-
### Internal Raw-Checkpoint Generation
|
| 290 |
-
|
| 291 |
-
For training-machine debugging and exact raw FSDP2 checkpoint recovery, the project still includes the upstream-style inference path:
|
| 292 |
-
|
| 293 |
-
- `simple_inference_engine.py`
|
| 294 |
-
- raw checkpoints from `LLM-OS-Models/KoHRM-Text-1.4B-raw-checkpoints`
|
| 295 |
-
- CUDA/FlashAttention-oriented execution
|
| 296 |
-
|
| 297 |
-
That path is mainly for internal continuation/evaluation, not the easiest Colab test.
|
| 298 |
-
|
| 299 |
-
|
| 300 |
-
### Training Data
|
| 301 |
-
|
| 302 |
-
Prepared data artifacts are uploaded to:
|
| 303 |
-
|
| 304 |
-
https://huggingface.co/datasets/LLM-OS-Models/KoHRM-Text-1.4B-prepared-data
|
| 305 |
-
|
| 306 |
-
The training objective is PrefixLM response-only loss. Instruction/prompt tokens are visible as context, while loss is applied to the response span.
|
| 307 |
-
|
| 308 |
-
Major prepared data groups:
|
| 309 |
-
|
| 310 |
-
| Dataset group | Tokens | Use |
|
| 311 |
-
|---|---:|---|
|
| 312 |
-
| `koterm_pretrain_mix_v1` | 711.3M | stage-0/stage0b |
|
| 313 |
-
| HRM cleaned fast-cap stage1/stage1b | 14.55B | HRM-style instruction pretraining |
|
| 314 |
-
| HRM cleaned full/no-cap stage2 | 14.55B | completed continuation |
|
| 315 |
-
| HRM cleaned full/no-cap extra stage2b | 14.55B | active continuation |
|
| 316 |
-
| Local terminal conversations | 9.39B | terminal/code/tool-heavy continuation |
|
| 317 |
-
| Korean tool/legal/wiki/finance mix | 3.02B | Korean domain and tool continuation |
|
| 318 |
-
| BCAI Finance Korean | 857.7M | Korean finance/domain data |
|
| 319 |
-
| Korean legal/admin task data | 629.0M | Korean legal/admin data |
|
| 320 |
-
| Korean Wikipedia | 462.5M | Korean general text |
|
| 321 |
-
| ToolBench train tool-call data | 127.0M | tool-call pretraining |
|
| 322 |
-
| SWE-ZERO + GLM reasoning subsets | 251.2M | code/reasoning data |
|
| 323 |
-
|
| 324 |
-
Evaluation-like datasets are excluded where identified, including ToolBench eval, Terminal Bench style evaluation data, and benchmark-oriented `chi-bench` data.
|
| 325 |
-
|
| 326 |
-
### Training Run
|
| 327 |
-
|
| 328 |
-
The current run uses staged continuation:
|
| 329 |
-
|
| 330 |
-
```text
|
| 331 |
-
stage0
|
| 332 |
-
-> stage0b
|
| 333 |
-
-> stage1
|
| 334 |
-
-> stage2
|
| 335 |
-
-> stage3
|
| 336 |
-
-> stage4
|
| 337 |
-
-> stage1b
|
| 338 |
-
-> stage2b
|
| 339 |
-
-> stage3b
|
| 340 |
-
-> stage4b
|
| 341 |
-
-> stage1c
|
| 342 |
-
-> stage2c
|
| 343 |
-
-> stage3c
|
| 344 |
-
-> stage4c
|
| 345 |
-
```
|
| 346 |
-
|
| 347 |
-
The checkpoint carries model weights, optimizer state, EMA weights, and recurrent carry state. `resume_step_offset` and `total_steps_override` are used so the learning-rate schedule follows the intended longer run instead of resetting at each stage.
|
| 348 |
-
|
| 349 |
-
As of 2026-05-27, `stage2b` is active. The continuation watcher is scheduled to launch `stage3b -> stage4b -> stage1c -> stage2c -> stage3c -> stage4c` after each completed checkpoint. The handoff reads the actual `epoch_1_info.json` `global_step` from each completed checkpoint before starting the next stage.
|
| 350 |
-
|
| 351 |
-
### Intended Use
|
| 352 |
-
|
| 353 |
-
This checkpoint is intended for:
|
| 354 |
-
|
| 355 |
-
- continued pretraining experiments
|
| 356 |
-
- Korean tokenizer and HRM-Text architecture experiments
|
| 357 |
-
- terminal/tool-call/code pretraining research
|
| 358 |
-
- checkpoint conversion and evaluation work
|
| 359 |
-
|
| 360 |
-
It is not yet intended as a finished assistant model.
|
| 361 |
-
|
| 362 |
-
### Limitations
|
| 363 |
-
|
| 364 |
-
- This is an intermediate checkpoint, not a final aligned instruct model.
|
| 365 |
-
- The full planned continuation has not finished.
|
| 366 |
-
- Final SFT and safety tuning have not been completed.
|
| 367 |
-
- Public benchmark scores for this new checkpoint are not final.
|
| 368 |
-
- Plain Transformers generation requires adding the custom `hrm_text` modeling wrapper or remote-code files.
|
| 369 |
-
- Tool-call JSON validity and terminal action safety must be evaluated before production use.
|
| 370 |
-
|
| 371 |
-
### Citation
|
| 372 |
-
|
| 373 |
-
This work builds on HRM-Text:
|
| 374 |
-
|
| 375 |
-
- Paper: https://arxiv.org/html/2605.20613
|
| 376 |
-
- Upstream code: https://github.com/sapientinc/HRM-Text
|
| 377 |
-
|
| 378 |
-
<a id="korean"></a>
|
| 379 |
-
|
| 380 |
-
## ํ๊ตญ์ด
|
| 381 |
-
|
| 382 |
-
`KoHRM-Text-1.4B`๋ `sapientinc/HRM-Text`์ PrefixLM ํ์ต ์คํ์ ๊ธฐ๋ฐ์ผ๋ก ์ฒ์๋ถํฐ ํ์ต ์ค์ธ ํ๊ตญ์ด/์์ด/์ฝ๋/ํฐ๋ฏธ๋/ํด์ฝ ๋ชจ๋ธ์
๋๋ค.
|
| 383 |
-
|
| 384 |
-
์ด ๋ชจ๋ธ์ `sapientinc/HRM-Text-1B`๋ฅผ ์ด์ด์ ํ์ธํ๋ํ ๋ชจ๋ธ์ด ์๋๋๋ค. ํ๊ตญ์ด์ ํฐ๋ฏธ๋/ํด์ฝ ํ์์ ๋ง์ถฐ ์๋ก ๋ง๋ 131K byte-level BPE tokenizer๋ฅผ ์ฌ์ฉํ๋ฉฐ, ๊ฐ์ค์น๋ scratch pretraining์ผ๋ก ํ์ตํฉ๋๋ค.
|
| 385 |
-
|
| 386 |
-
### ํ์ฌ ์ํ
|
| 387 |
-
|
| 388 |
-
์ด ์ ์ฅ์๋ ์ต์ ๊ณต๊ฐ ๋ณํ๋ณธ์ ๊ณ์ ๋ฎ์ด์ฐ๋ rolling latest model repo์
๋๋ค. ํ์ต์ ์์ง ์งํ ์ค์
๋๋ค.
|
| 389 |
-
|
| 390 |
-
- ๋ฉ์ธ ๋ชจ๋ธ repo: `LLM-OS-Models/KoHRM-Text-1.4B`
|
| 391 |
-
- ํ์ฌ ๊ณต๊ฐ ํ์ผ: `model.safetensors`, `config.json`, tokenizer ํ์ผ, `README.md`
|
| 392 |
-
- raw FSDP2 resume checkpoint: `LLM-OS-Models/KoHRM-Text-1.4B-raw-checkpoints`
|
| 393 |
-
- prepared data: `LLM-OS-Models/KoHRM-Text-1.4B-prepared-data`
|
| 394 |
-
- ํ๋ก์ ํธ ์ฝ๋: https://github.com/LLM-OS-Models/KoHRM-text
|
| 395 |
-
- ์๋ณธ HRM-Text ์ฝ๋: https://github.com/sapientinc/HRM-Text
|
| 396 |
-
- HRM-Text ๋
ผ๋ฌธ: https://arxiv.org/html/2605.20613
|
| 397 |
-
- tokenizer repo: `LLM-OS-Models/HRM-Text-Ko-Terminal-Tokenizer-131K`
|
| 398 |
-
|
| 399 |
-
์ต์ ๊ณต๊ฐ weight๋ฅผ ํ
์คํธํ๋ ค๋ฉด `revision="main"`์ผ๋ก ๋ค์ด๋ก๋ํ๋ฉด ๋ฉ๋๋ค. ํ์ต ์ค 10,000 step ๋จ์๋ก ์ checkpoint๊ฐ ๋ณํ๋์ด ์ฌ๋ผ์ค๋ฉด ๊ฐ์ ํ์ผ๋ช
์ด ์ต์ EMA `safetensors`๋ก ๊ฐฑ์ ๋ฉ๋๋ค.
|
| 400 |
-
|
| 401 |
-
### ํ์ต ๋ฐฉ์ ํ๋์ ๋ณด๊ธฐ
|
| 402 |
-
|
| 403 |
-
KoHRM-Text๋ **scratch instruction pretraining**์ผ๋ก ๋ณด๋ ๊ฒ์ด ๊ฐ์ฅ ์ ํํฉ๋๋ค.
|
| 404 |
-
|
| 405 |
-
์ผ๋ฐ์ ์ธ raw-text causal LM ์ฌ์ ํ์ต๋ ์๋๊ณ , ์ด๋ฏธ ์์ฑ๋ base model ์์ ์งง๊ฒ ์น๋ SFT๋ง๋ ์๋๋๋ค.
|
| 406 |
-
|
| 407 |
-
```text
|
| 408 |
-
raw data -> tokenizer -> V1Dataset -> PrefixLM batches
|
| 409 |
-
-> HRM H/L recurrence -> LM head -> response-only loss
|
| 410 |
-
```
|
| 411 |
-
|
| 412 |
-
์
๋ ฅ ์ปจํ
์คํธ๋ PrefixLM prefix๋ก ์ฒ๋ฆฌํฉ๋๋ค.
|
| 413 |
-
|
| 414 |
-
```text
|
| 415 |
-
instruction / prefix: ์๋ฐฉํฅ attention, loss ์์
|
| 416 |
-
response: causal attention, response-only CE loss
|
| 417 |
```
|
| 418 |
|
| 419 |
-
|
| 420 |
-
|
| 421 |
-
```text
|
| 422 |
-
H module: ๋๋ฆฌ๊ฒ ๋ณํ๋ ์ ๋ต state
|
| 423 |
-
L module: ๋น ๋ฅด๊ฒ ๋ณํ๋ ์คํ state
|
| 424 |
-
schedule: H2L3 recurrent computation
|
| 425 |
-
```
|
| 426 |
-
|
| 427 |
-
ํ์ต ๋ฐฉ์, ์ํคํ
์ฒ, PT/SFT ์ฐจ์ด, staged continuation, checkpoint ์ด๋ฆ์ ์ฝ๊ฒ ํ์ด ์ด ์ ์ฒด ์ค๋ช
์ ํ๋ก์ ํธ ๋ฌธ์๋ฅผ ๊ธฐ์ค์ผ๋ก ๋ณด๋ฉด ๋ฉ๋๋ค.
|
| 428 |
-
|
| 429 |
-
[MODEL_TRAINING_ARCHITECTURE_GUIDE_2026-05-28.md](MODEL_TRAINING_ARCHITECTURE_GUIDE_2026-05-28.md) in https://github.com/LLM-OS-Models/KoHRM-text
|
| 430 |
-
|
| 431 |
-
### ์ค์ํ ํธํ์ฑ ์๋ด
|
| 432 |
-
|
| 433 |
-
ํ์ฌ ๊ณต๊ฐ repo์๋ ๋ณํ๋ model weight์ tokenizer๊ฐ ์์ง๋ง, ์์ง Hugging Face `trust_remote_code`์ฉ `HrmTextForCausalLM` ๊ตฌํ ํ์ผ์ ํฌํจ๋์ด ์์ง ์์ต๋๋ค.
|
| 434 |
-
|
| 435 |
-
ํ์ฌ ๋ฐ๋ก ๊ฐ๋ฅํ ๊ฒ:
|
| 436 |
-
|
| 437 |
-
- ์ต์ ๊ณต๊ฐ weight ๋ค์ด๋ก๋
|
| 438 |
-
- `tokenizers.Tokenizer.from_file("tokenizer.json")`๋ก tokenizer ๋ก๋
|
| 439 |
-
- `config.json` ํ์ธ
|
| 440 |
-
- CPU ๋๋ Colab T4์์ `model.safetensors` ๋ฌด๊ฒฐ์ฑ ํ์ธ
|
| 441 |
-
|
| 442 |
-
์์ง ์ผ๋ฐ Transformers์์ ๋ฐ๋ก ์ ๋๋ ๊ฒ:
|
| 443 |
-
|
| 444 |
-
- `AutoModelForCausalLM.from_pretrained("LLM-OS-Models/KoHRM-Text-1.4B")`
|
| 445 |
-
- ์ด repo๋ง์ผ๋ก one-line text generation ์คํ
|
| 446 |
-
|
| 447 |
-
์ด์ ๋ `model_type: "hrm_text"`๊ฐ custom HRM-Text architecture์ด๊ธฐ ๋๋ฌธ์
๋๋ค. ๊ณต๊ฐ generation์ ํ๋ ค๋ฉด ์ด model repo์ `HrmTextForCausalLM` remote-code wrapper๊ฐ ์ถ๊ฐ๋์ด์ผ ํฉ๋๋ค.
|
| 448 |
-
|
| 449 |
-
### ๋ชจ๋ธ ์์ธ
|
| 450 |
-
|
| 451 |
-
| ํญ๋ชฉ | ๊ฐ |
|
| 452 |
-
|---|---:|
|
| 453 |
-
| ๋ชจ๋ธ ID | `LLM-OS-Models/KoHRM-Text-1.4B` |
|
| 454 |
-
| ํ์ค ์ด๋ฆ | `KoHRM-Text-1.4B` |
|
| 455 |
-
| ํ์ต ์ถ๋ฐ์ | scratch |
|
| 456 |
-
| ์ํคํ
์ฒ ๊ณ์ด | HRM-Text PrefixLM |
|
| 457 |
-
| ์ํคํ
์ฒ ํฌ๊ธฐ | `XL` |
|
| 458 |
-
| ํ๋ผ๋ฏธํฐ | 1,384,120,320 |
|
| 459 |
-
| ์ปจํ
์คํธ ๊ธธ์ด | 4,096 tokens |
|
| 460 |
-
| ํ์ต dtype | bfloat16 |
|
| 461 |
-
| ๊ณต๊ฐ ๋ณํ๋ณธ dtype | bfloat16 EMA `safetensors` |
|
| 462 |
-
| tokenizer | byte-level BPE, NFC normalization |
|
| 463 |
-
| vocabulary size | 131,072 |
|
| 464 |
-
| objective | PrefixLM response-only loss |
|
| 465 |
-
| optimizer | HRM-Text์ Adam-atan2 |
|
| 466 |
-
| EMA | 0.9999 |
|
| 467 |
-
|
| 468 |
-
๋ณํ๋ config ์ฃผ์ ๊ฐ:
|
| 469 |
|
| 470 |
```json
|
| 471 |
{
|
| 472 |
-
"
|
| 473 |
-
"
|
| 474 |
-
"
|
| 475 |
-
|
| 476 |
-
|
| 477 |
-
"
|
| 478 |
-
"max_position_embeddings": 4096,
|
| 479 |
-
"prefix_lm": true
|
| 480 |
}
|
| 481 |
```
|
| 482 |
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
ํ์ฌ run์ ๋
ผ๋ฌธ recipe๋ณด๋ค ๋ ์ค๋ ๊ฑธ๋ฆด ์ ์์ต๋๋ค. ์ค์ ์ด ์์ ํ ๊ฐ์ง ์๊ธฐ ๋๋ฌธ์
๋๋ค.
|
| 486 |
-
|
| 487 |
-
- ๋
ผ๋ฌธ ๊ธฐ์ค์ 16 x H100์ด๊ณ , ํ์ฌ run์ 8 x H200์
๋๋ค.
|
| 488 |
-
- KoHRM์ ์๋ณธ 65K tokenizer๋ณด๋ค ํฐ 131K tokenizer vocab์ ์๋๋ค.
|
| 489 |
-
- ๊ณต๊ฐ KoHRM ํฌ๊ธฐ๋ ์ฝ 1.38B parameters์
๋๋ค.
|
| 490 |
-
- ์์ ์ฅ๊ธฐ run batch๋ OOM probe ์ดํ `180,224` tokens/step์ผ๋ก ์ก์์ต๋๋ค. ๋ ํฐ batch๋ ์ด๋ฐ์ ๊ฐ๋ฅํด ๋ณด์ฌ๋ ์ฅ๊ธฐ ์์ ์ฑ์ด ๋จ์ด์ก์ต๋๋ค.
|
| 491 |
-
- ํ๊ตญ์ด, ํฐ๋ฏธ๋, ํด์ฝ, ๋ฒ๋ฅ , ๊ธ์ต, ์ํค, HRM-cleaned ๋ฐ๋ณต stage๊ฐ ์ถ๊ฐ๋์ต๋๋ค.
|
| 492 |
-
|
| 493 |
-
์ด๊ฒ์ด ์๋์ผ๋ก ๋ชจ๋ benchmark ์ ์ ์์น์ ๋ณด์ฅํ์ง๋ ์์ต๋๋ค. ๋ค๋ง ํ๊ตญ์ด ํ ํฌ๋์ด์ ํจ์จ, ํ๊ตญ์ด ๋ฒ๋ฅ /๊ธ์ต/์ํค coverage, ํฐ๋ฏธ๋ trajectory, tool-call formatting, code-oriented behavior ์ชฝ์ ์๋ณธ ์์ด/general checkpoint๋ณด๋ค ์ข์์ง ๊ฐ๋ฅ์ฑ์ด ์์ต๋๋ค. ์ต์ข
์ฃผ์ฅ์ continuation๊ณผ SFT๊ฐ ๋๋ ๋ค ํ๊ฐ๋ก ํ์ธํด์ผ ํฉ๋๋ค.
|
| 494 |
-
|
| 495 |
-
### ํ ํฌ๋์ด์
|
| 496 |
-
|
| 497 |
-
ํ ํฌ๋์ด์ ๋ ํ๊ตญ์ด, ์์ด, ์ฝ๋, shell/terminal ํ
์คํธ, JSON/tool-call ํ์์ ๊ณ ๋ คํด์ ๋ง๋ค์์ต๋๋ค. ์์ฃผ ์ฐ๋ chat/tool special token์ ๊ฐ๋ฅํ ํ ์์ ์ ์ธ ๋จ์ผ token์ผ๋ก ์ ์งํฉ๋๋ค.
|
| 498 |
-
|
| 499 |
-
| ์ํ ์ข
๋ฅ | chars/token |
|
| 500 |
-
|---|---:|
|
| 501 |
-
| ํ๊ตญ์ด ์ผ๋ฐ | 2.60 |
|
| 502 |
-
| ํ๊ตญ์ด ๋ฒ๋ฅ | 2.36 |
|
| 503 |
-
| ํ๊ตญ์ด ํฐ๋ฏธ๋ ์ง์ | 2.18 |
|
| 504 |
-
| shell command | 2.68 |
|
| 505 |
-
| tool-call JSON | 3.32 |
|
| 506 |
-
| Python code | 3.37 |
|
| 507 |
-
| ์์ด | 4.40 |
|
| 508 |
-
|
| 509 |
-
ํฌ๋งท token:
|
| 510 |
-
|
| 511 |
-
```text
|
| 512 |
-
<|im_start|> instruction ์์
|
| 513 |
-
<|im_end|> instruction ์ข
๋ฃ
|
| 514 |
-
<|box_end|> response/end marker
|
| 515 |
-
<|object_ref_start|> direct condition
|
| 516 |
-
<|object_ref_end|> chain-of-thought style condition
|
| 517 |
-
<|quad_start|> noisy condition
|
| 518 |
-
<|quad_end|> synthetic condition
|
| 519 |
-
```
|
| 520 |
-
|
| 521 |
-
ํ๋ก์ ํธ ๋ด๋ถ inference code๊ฐ ์ฐ๋ prompt ํ์:
|
| 522 |
|
| 523 |
-
```
|
| 524 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 525 |
```
|
| 526 |
|
| 527 |
-
|
| 528 |
-
|
| 529 |
-
|
| 530 |
-
|
| 531 |
-
https://github.com/LLM-OS-Models/KoHRM-text/blob/main/notebooks/KoHRM_Text_1_4B_Colab_T4_Smoke_Test.ipynb
|
| 532 |
-
|
| 533 |
-
์ด ๋
ธํธ๋ถ์ Colab T4์์ ์ต์ ๊ณต๊ฐ ํ์ผ์ ๋ค์ด๋ก๋ํ๊ณ ์งง์ ์์ฑ์ ์ง์ ์คํํฉ๋๋ค.
|
| 534 |
-
|
| 535 |
-
์ผ๋ถ Colab ํ๊ฒฝ์์ `transformers`๊ฐ `torchvision::nms` import ์ค๋ฅ๋ฅผ ๋ด๊ฑฐ๋ custom architecture๋ฅผ ๋ชป ์ฐพ๋ ๋ฌธ์ ๊ฐ ์๊ธธ ์ ์์ผ๋ฏ๋ก, ์ด ๋
ธํธ๋ถ์ `AutoTokenizer`์ `AutoModelForCausalLM`์ ์ฐ์ง ์์ต๋๋ค. ๋์ ์๋ ๊ฒฝ๋ก๋ฅผ ์ฌ์ฉํฉ๋๋ค.
|
| 536 |
-
|
| 537 |
-
- `tokenizers.Tokenizer.from_file("tokenizer.json")`
|
| 538 |
-
- `safetensors.torch.load_file("model.safetensors")`
|
| 539 |
-
- HRM-Text ๊ตฌ์กฐ๋ฅผ ์ง์ ๊ตฌํํ `kohrm_colab_generate.py`
|
| 540 |
-
|
| 541 |
-
```python
|
| 542 |
-
!pip -q install -U huggingface_hub hf_transfer tokenizers safetensors
|
| 543 |
-
```
|
| 544 |
-
|
| 545 |
-
```python
|
| 546 |
-
from pathlib import Path
|
| 547 |
-
import json
|
| 548 |
-
import importlib.util
|
| 549 |
-
import sys
|
| 550 |
-
from huggingface_hub import snapshot_download
|
| 551 |
-
from tokenizers import Tokenizer
|
| 552 |
-
|
| 553 |
-
repo_id = "LLM-OS-Models/KoHRM-Text-1.4B"
|
| 554 |
-
|
| 555 |
-
repo_dir = Path(snapshot_download(
|
| 556 |
-
repo_id,
|
| 557 |
-
revision="main",
|
| 558 |
-
allow_patterns=[
|
| 559 |
-
"README.md",
|
| 560 |
-
"config.json",
|
| 561 |
-
"tokenizer.json",
|
| 562 |
-
"tokenizer_config.json",
|
| 563 |
-
"special_tokens_map.json",
|
| 564 |
-
"model.safetensors",
|
| 565 |
-
"kohrm_colab_generate.py",
|
| 566 |
-
],
|
| 567 |
-
))
|
| 568 |
-
|
| 569 |
-
print("Downloaded to:", repo_dir)
|
| 570 |
-
config = json.loads((repo_dir / "config.json").read_text())
|
| 571 |
-
print("model_type:", config["model_type"])
|
| 572 |
-
print("hidden_size:", config["hidden_size"])
|
| 573 |
-
print("vocab_size:", config["vocab_size"])
|
| 574 |
-
print("context:", config["max_position_embeddings"])
|
| 575 |
-
|
| 576 |
-
tokenizer = Tokenizer.from_file(str(repo_dir / "tokenizer.json"))
|
| 577 |
-
wrapped = "<|im_start|><|object_ref_start|>ํ๊ตญ์ด๋ก ํ์ฌ ๋๋ ํฐ๋ฆฌ์์ ๊ฐ์ฅ ํฐ ํ์ผ 10๊ฐ๋ฅผ ์ฐพ๋ ๋ช
๋ น์ ์๋ ค์ฃผ์ธ์.<|im_end|>"
|
| 578 |
-
ids = tokenizer.encode(wrapped).ids
|
| 579 |
-
print("prompt tokens:", len(ids))
|
| 580 |
-
print("first token ids:", ids[:20])
|
| 581 |
-
|
| 582 |
-
spec = importlib.util.spec_from_file_location(
|
| 583 |
-
"kohrm_colab_generate",
|
| 584 |
-
repo_dir / "kohrm_colab_generate.py",
|
| 585 |
-
)
|
| 586 |
-
kohrm = importlib.util.module_from_spec(spec)
|
| 587 |
-
sys.modules["kohrm_colab_generate"] = kohrm
|
| 588 |
-
spec.loader.exec_module(kohrm)
|
| 589 |
-
|
| 590 |
-
output = kohrm.generate_text(
|
| 591 |
-
repo_dir,
|
| 592 |
-
"ํ๊ตญ์ด๋ก ํ์ฌ ๋๋ ํฐ๋ฆฌ์์ ๊ฐ์ฅ ํฐ ํ์ผ 10๊ฐ๋ฅผ ์ฐพ๋ bash ๋ช
๋ น์ ์๋ ค์ฃผ์ธ์. ๋ช
๋ น๋ง ๊ฐ๋จํ ๋ตํ์ธ์.",
|
| 593 |
-
max_new_tokens=64,
|
| 594 |
-
max_seq_len=512,
|
| 595 |
-
temperature=0.0,
|
| 596 |
-
)
|
| 597 |
-
print(output)
|
| 598 |
-
```
|
| 599 |
-
|
| 600 |
-
์ ์ ๊ฒฐ๊ณผ:
|
| 601 |
-
|
| 602 |
-
- `model_type`์ `hrm_text`์
๋๋ค.
|
| 603 |
-
- `vocab_size`๋ `131072`์
๋๋ค.
|
| 604 |
-
- helper๊ฐ 1.38B ๊ณต๊ฐ `model.safetensors` ๋ณํ๋ณธ์ ๋ก๋ํฉ๋๋ค.
|
| 605 |
-
- Colab T4์์๋ fp16 PyTorch scaled-dot-product attention์ผ๋ก ์์ฑํฉ๋๋ค.
|
| 606 |
-
- ์ฒซ ์คํ์ 2.8 GiB๊ธ weight ๋ค์ด๋ก๋์ ๋ก๋ ๋๋ฌธ์ ๋ช ๋ถ ๊ฑธ๋ฆด ์ ์์ต๋๋ค.
|
| 607 |
-
|
| 608 |
-
helper๊ฐ ์ฐ๋ prompt ํ์:
|
| 609 |
-
|
| 610 |
-
```text
|
| 611 |
-
<|im_start|><|object_ref_start|>PROMPT<|im_end|>
|
| 612 |
-
```
|
| 613 |
-
|
| 614 |
-
์ผ๋ฐ `AutoModelForCausalLM.generate()`๋ ์์ง ์ง์ ๊ฒฝ๋ก๊ฐ ์๋๋๋ค. ์ด ๋ชจ๋ธ์ custom `hrm_text` architecture์ด๋ฏ๋ก, ์ผ๋ฐ Transformers generation์ ์ถํ `trust_remote_code` wrapper๊ฐ ์ถ๊ฐ๋ ๋ค ์ง์ํ๋ ๊ฒ์ด ๋ง์ต๋๋ค. ์ง๊ธ ๊ณต๊ฐ `model.safetensors`๋ก ๋ฐ๋ก ์์ฑํ๋ ค๋ฉด ์ ๋
ธํธ๋ถ/helper๋ฅผ ์ฐ๋ฉด ๋ฉ๋๋ค.
|
| 615 |
-
|
| 616 |
-
### ๋ด๋ถ raw-checkpoint ์์ฑ
|
| 617 |
-
|
| 618 |
-
ํ์ต ๋จธ์ ์์ ๋๋ฒ๊น
ํ๊ฑฐ๋ raw FSDP2 checkpoint๋ฅผ ์ ํํ ๋ณต๊ตฌํด์ ํ๊ฐํ ๋๋ upstream ์คํ์ผ inference ๊ฒฝ๋ก๋ ์ ์งํฉ๋๋ค.
|
| 619 |
-
|
| 620 |
-
- `simple_inference_engine.py`
|
| 621 |
-
- `LLM-OS-Models/KoHRM-Text-1.4B-raw-checkpoints`์ raw checkpoints
|
| 622 |
-
- CUDA/FlashAttention ์ค์ฌ ์คํ
|
| 623 |
-
|
| 624 |
-
์ด ๊ฒฝ๋ก๋ ๋ด๋ถ continuation/evaluation์ฉ์ ๊ฐ๊น๊ณ , Colab์์ ๊ฐ์ฅ ์ฝ๊ฒ ํ์ธํ๋ ค๋ฉด ์ ๊ณต๊ฐ `model.safetensors` helper๋ฅผ ์ฐ๋ ๊ฒ์ด ๋ซ์ต๋๋ค.
|
| 625 |
-
|
| 626 |
-
### ํ์ต ๋ฐ์ดํฐ
|
| 627 |
-
|
| 628 |
-
prepared data๋ ์๋ dataset repo์ ์
๋ก๋ํฉ๋๋ค.
|
| 629 |
-
|
| 630 |
-
https://huggingface.co/datasets/LLM-OS-Models/KoHRM-Text-1.4B-prepared-data
|
| 631 |
-
|
| 632 |
-
ํ์ต objective๋ PrefixLM response-only loss์
๋๋ค. instruction/prompt token์ context๋ก ๋ณด๊ณ , loss๋ response span์๋ง ์ ์ฉํฉ๋๋ค.
|
| 633 |
-
|
| 634 |
-
์ฃผ์ prepared data group:
|
| 635 |
-
|
| 636 |
-
| ๋ฐ์ดํฐ ๊ทธ๋ฃน | Tokens | ์ฉ๋ |
|
| 637 |
-
|---|---:|---|
|
| 638 |
-
| `koterm_pretrain_mix_v1` | 711.3M | stage-0/stage0b |
|
| 639 |
-
| HRM cleaned fast-cap stage1/stage1b | 14.55B | HRM-style instruction pretraining |
|
| 640 |
-
| HRM cleaned full/no-cap stage2 | 14.55B | ์๋ฃ๋ continuation |
|
| 641 |
-
| HRM cleaned full/no-cap extra stage2b | 14.55B | ์งํ ์ค์ธ continuation |
|
| 642 |
-
| local terminal conversations | 9.39B | terminal/code/tool-heavy continuation |
|
| 643 |
-
| Korean tool/legal/wiki/finance mix | 3.02B | ํ๊ตญ์ด domain/tool continuation |
|
| 644 |
-
| BCAI Finance Korean | 857.7M | ํ๊ตญ์ด ๊ธ์ต/domain data |
|
| 645 |
-
| Korean legal/admin task data | 629.0M | ํ๊ตญ์ด ๋ฒ๋ฅ /ํ์ data |
|
| 646 |
-
| Korean Wikipedia | 462.5M | ํ๊ตญ์ด ์ผ๋ฐ ํ
์คํธ |
|
| 647 |
-
| ToolBench train tool-call data | 127.0M | tool-call pretraining |
|
| 648 |
-
| SWE-ZERO + GLM reasoning subsets | 251.2M | code/reasoning data |
|
| 649 |
-
|
| 650 |
-
ํ๊ฐ ์ฑ๊ฒฉ ๋ฐ์ดํฐ๋ ํ์ธ๋๋ ๋ฒ์์์ train์์ ์ ์ธํฉ๋๋ค. ์์๋ ToolBench eval, Terminal Bench ๊ณ์ด ํ๊ฐ ๋ฐ์ดํฐ, benchmark ์ฑ๊ฒฉ์ `chi-bench`์
๋๋ค.
|
| 651 |
-
|
| 652 |
-
### ํ์ต ์งํ
|
| 653 |
-
|
| 654 |
-
ํ์ฌ run์ staged continuation ๋ฐฉ์์
๋๋ค.
|
| 655 |
-
|
| 656 |
-
```text
|
| 657 |
-
stage0
|
| 658 |
-
-> stage0b
|
| 659 |
-
-> stage1
|
| 660 |
-
-> stage2
|
| 661 |
-
-> stage3
|
| 662 |
-
-> stage4
|
| 663 |
-
-> stage1b
|
| 664 |
-
-> stage2b
|
| 665 |
-
-> stage3b
|
| 666 |
-
-> stage4b
|
| 667 |
-
-> stage1c
|
| 668 |
-
-> stage2c
|
| 669 |
-
-> stage3c
|
| 670 |
-
-> stage4c
|
| 671 |
-
```
|
| 672 |
-
|
| 673 |
-
checkpoint๋ model weights, optimizer state, EMA weights, recurrent carry state๋ฅผ ์ด์ด๊ฐ๋๋ค. `resume_step_offset`๊ณผ `total_steps_override`๋ฅผ ์จ์ stage๋ง๋ค learning-rate schedule์ด ๋ฆฌ์
๋์ง ์๊ณ ๊ธด pretraining run์ฒ๋ผ ์ด์ด์ง๊ฒ ํฉ๋๋ค.
|
| 674 |
-
|
| 675 |
-
2026-05-27 ๊ธฐ์ค `stage2b`๊ฐ ์งํ ์ค์
๋๋ค. continuation watcher๊ฐ ์ดํ `stage3b -> stage4b -> stage1c -> stage2c -> stage3c -> stage4c`๋ฅผ ์ด์ด์ ์คํํ๋๋ก ์์ฝ๋์ด ์์ต๋๋ค. handoff๋ ๊ฐ stage์ ์ค์ `epoch_1_info.json` `global_step`์ ์ฝ๊ณ ๋ค์ stage๋ฅผ ์์ํฉ๋๋ค.
|
| 676 |
-
|
| 677 |
-
### ์ฌ์ฉ ๋ชฉ์
|
| 678 |
-
|
| 679 |
-
์ด checkpoint๋ ๋ค์ ๋ชฉ์ ์ ์ ํฉํฉ๋๋ค.
|
| 680 |
-
|
| 681 |
-
- continued pretraining ์คํ
|
| 682 |
-
- ํ๊ตญ์ด tokenizer ๋ฐ HRM-Text architecture ์คํ
|
| 683 |
-
- terminal/tool-call/code pretraining ์ฐ๊ตฌ
|
| 684 |
-
- checkpoint conversion ๋ฐ evaluation ์์
|
| 685 |
-
|
| 686 |
-
์์ง ์์ฑ๋ assistant model์ ์๋๋๋ค.
|
| 687 |
-
|
| 688 |
-
### ์ ํ ์ฌํญ
|
| 689 |
|
| 690 |
-
|
| 691 |
-
- ์ ์ฒด planned continuation์ด ์์ง ๋๋์ง ์์์ต๋๋ค.
|
| 692 |
-
- ์ต์ข
SFT์ safety tuning์ด ์์ง ๋๋์ง ์์์ต๋๋ค.
|
| 693 |
-
- ์ checkpoint์ public benchmark score๋ ์์ง final์ด ์๋๋๋ค.
|
| 694 |
-
- ์ผ๋ฐ Transformers generation์ custom `hrm_text` modeling wrapper ๋๋ remote-code file์ด ์ถ๊ฐ๋์ด์ผ ๊ฐ๋ฅํฉ๋๋ค.
|
| 695 |
-
- tool-call JSON ์ ํจ์ฑ๊ณผ terminal action safety๋ ์ค์ ์ฌ์ฉ ์ ์ ๋ณ๋ ํ๊ฐ๊ฐ ํ์ํฉ๋๋ค.
|
| 696 |
|
| 697 |
-
|
|
|
|
|
|
|
| 698 |
|
| 699 |
-
|
| 700 |
|
| 701 |
-
-
|
| 702 |
-
-
|
|
|
|
|
|
| 1 |
---
|
|
|
|
| 2 |
language:
|
|
|
|
| 3 |
- en
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
- ko
|
| 5 |
+
library_name: transformers
|
| 6 |
+
pipeline_tag: text-generation
|
| 7 |
tags:
|
|
|
|
|
|
|
| 8 |
- terminal
|
| 9 |
+
- sft
|
| 10 |
+
- vllm
|
| 11 |
+
- tb2-lite
|
| 12 |
+
- evaluation-pending
|
| 13 |
+
base_model: unknown
|
|
|
|
| 14 |
---
|
| 15 |
|
| 16 |
+
# LLM-OS-Models/KoHRM-Text-1.4B-Epoch2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
+
ํฐ๋ฏธ๋ ์์
์๋ํ๋ฅผ ์ํ Terminal SFT ๋ชจ๋ธ์
๋๋ค. ์
๋ ฅ๋ ์์
/์ด์ ํฐ๋ฏธ๋ ์ํ๋ฅผ ๋ณด๊ณ ๋ค์์ ์คํํ ๋ช
๋ น์ JSON ํํ๋ก ์์ฑํ๋ ์ฉ๋๋ก ํ์ตํ์ต๋๋ค.
|
| 19 |
|
| 20 |
+
## ๋ชจ๋ธ ์์ฝ
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
+
- Base model: `unknown`
|
| 23 |
+
- Training setup: `Terminal SFT`
|
| 24 |
+
- Model card snapshot: `2026-05-29 01:06:51 UTC`
|
| 25 |
+
- Corrected TB2-lite evaluated results currently indexed: `56`
|
| 26 |
+
- Corrected TB2-lite score: `pending / not matched in current result directory`
|
| 27 |
|
| 28 |
+
## Quickstart
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
|
| 30 |
+
์ค์น์ ๋ก๊ทธ์ธ:
|
| 31 |
|
| 32 |
+
```bash
|
| 33 |
+
pip install -U vllm transformers huggingface_hub
|
| 34 |
+
huggingface-cli login
|
| 35 |
```
|
| 36 |
|
| 37 |
+
๊ด๋ จ ์ฝ๋:
|
| 38 |
|
| 39 |
+
- GitHub: https://github.com/LLM-OS-Models/Terminal
|
| 40 |
+
- vLLM ํ๊ฐ ์คํ: `tb2_lite/scripts/replay_eval.py`
|
| 41 |
+
- chat template/fallback ์์ฑ: `tb2_lite/scripts/prompt_builder.py`
|
| 42 |
+
- JSON/command ์ฑ์ : `tb2_lite/scripts/replay_metrics.py`
|
| 43 |
|
| 44 |
+
vLLM ์ง์ ์คํ ์์. ํ๊ฐ ์ฝ๋์ ๋์ผํ๊ฒ chat template์ ์ฐ์ ์ฌ์ฉํ๊ณ , template์ด ์์ผ๋ฉด ChatML/Gemma fallback์ ์ฌ์ฉํฉ๋๋ค.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
```python
|
| 47 |
+
from transformers import AutoTokenizer
|
| 48 |
+
from vllm import LLM, SamplingParams
|
| 49 |
+
|
| 50 |
+
model_id = "LLM-OS-Models/KoHRM-Text-1.4B-Epoch2"
|
| 51 |
+
tp = 1
|
| 52 |
+
|
| 53 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
|
| 54 |
+
llm = LLM(
|
| 55 |
+
model=model_id,
|
| 56 |
+
tokenizer=model_id,
|
| 57 |
+
trust_remote_code=True,
|
| 58 |
+
dtype="bfloat16",
|
| 59 |
+
tensor_parallel_size=tp,
|
| 60 |
+
max_model_len=49152,
|
| 61 |
+
gpu_memory_utilization=0.92,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
)
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
+
messages = [
|
| 65 |
+
{"role": "system", "content": "You are a terminal automation assistant. Return JSON only."},
|
| 66 |
+
{"role": "user", "content": "Inspect the current directory and list Python files."},
|
| 67 |
+
]
|
| 68 |
+
|
| 69 |
+
def render_chatml(messages):
|
| 70 |
+
parts = []
|
| 71 |
+
for message in messages:
|
| 72 |
+
role = "assistant" if message["role"] == "assistant" else message["role"]
|
| 73 |
+
if role == "tool":
|
| 74 |
+
role = "user"
|
| 75 |
+
parts.append(f"<|im_start|>{role}\n{message['content']}<|im_end|>\n")
|
| 76 |
+
parts.append("<|im_start|>assistant\n")
|
| 77 |
+
return "".join(parts)
|
| 78 |
+
|
| 79 |
+
def render_gemma4_turn(messages, empty_thought_channel=False):
|
| 80 |
+
parts = ["<bos>"]
|
| 81 |
+
for message in messages:
|
| 82 |
+
role = "model" if message["role"] == "assistant" else message["role"]
|
| 83 |
+
if role == "tool":
|
| 84 |
+
role = "user"
|
| 85 |
+
parts.append(f"<|turn>{role}\n{message['content'].strip()}<turn|>\n")
|
| 86 |
+
parts.append("<|turn>model\n")
|
| 87 |
+
if empty_thought_channel:
|
| 88 |
+
parts.append("<|channel>thought\n<channel|>")
|
| 89 |
+
return "".join(parts)
|
| 90 |
+
|
| 91 |
+
def render_prompt(model_id, tokenizer, messages):
|
| 92 |
+
model_key = model_id.lower()
|
| 93 |
+
if "gemma-4" in model_key:
|
| 94 |
+
try:
|
| 95 |
+
return tokenizer.apply_chat_template(
|
| 96 |
+
messages,
|
| 97 |
+
tokenize=False,
|
| 98 |
+
add_generation_prompt=True,
|
| 99 |
+
enable_thinking=False,
|
| 100 |
+
)
|
| 101 |
+
except Exception:
|
| 102 |
+
return render_gemma4_turn(
|
| 103 |
+
messages,
|
| 104 |
+
empty_thought_channel=("26b" in model_key or "31b" in model_key),
|
| 105 |
+
)
|
| 106 |
+
try:
|
| 107 |
+
return tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
| 108 |
+
except Exception:
|
| 109 |
+
return render_chatml(messages)
|
| 110 |
+
|
| 111 |
+
prompt = render_prompt(model_id, tokenizer, messages)
|
| 112 |
+
sampling = SamplingParams(
|
| 113 |
temperature=0.0,
|
| 114 |
+
top_p=1.0,
|
| 115 |
+
max_tokens=1024,
|
| 116 |
+
repetition_penalty=1.0,
|
| 117 |
)
|
| 118 |
+
outputs = llm.generate([prompt], sampling_params=sampling)
|
| 119 |
+
print(outputs[0].outputs[0].text)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 120 |
```
|
| 121 |
|
| 122 |
+
๊ถ์ฅ ์ถ๋ ฅ ํ์:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
```json
|
| 125 |
{
|
| 126 |
+
"analysis": "brief reasoning about the next terminal action",
|
| 127 |
+
"plan": "short execution plan",
|
| 128 |
+
"commands": [
|
| 129 |
+
{"keystrokes": "ls -la\n", "duration": 0.1}
|
| 130 |
+
],
|
| 131 |
+
"task_complete": false
|
|
|
|
|
|
|
| 132 |
}
|
| 133 |
```
|
| 134 |
|
| 135 |
+
ํ๊ฐ์ ๋์ผํ replay ๋ช
๋ น:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 136 |
|
| 137 |
+
```bash
|
| 138 |
+
python tb2_lite/scripts/replay_eval.py \
|
| 139 |
+
--model LLM-OS-Models/KoHRM-Text-1.4B-Epoch2 \
|
| 140 |
+
--model-short LLM-OS-Models__KoHRM-Text-1.4B-Epoch2 \
|
| 141 |
+
--eval-path tb2_lite/data/replay_full.jsonl \
|
| 142 |
+
--output-dir /home/work/.data/tb2_lite_eval/corrected_readme_models_vllm \
|
| 143 |
+
--dtype bfloat16 \
|
| 144 |
+
--tp 1 \
|
| 145 |
+
--max-model-len 49152 \
|
| 146 |
+
--max-tokens 1024 \
|
| 147 |
+
--temperature 0.0 \
|
| 148 |
+
--top-p 1.0 \
|
| 149 |
+
--gpu-memory-utilization 0.92 \
|
| 150 |
+
--language-model-only
|
| 151 |
```
|
| 152 |
|
| 153 |
+
- ๊ธฐ๋ณธ ๊ถ์ฅ tensor parallel: `1`. OOM์ด๋ฉด `--tp`์ `tensor_parallel_size`๋ฅผ 2/4/8๋ก ์ฌ๋ฆฌ์ธ์.
|
| 154 |
+
- corrected TB2-lite ํ๊ฐ๋ `temperature=0.0`, `top_p=1.0`, `max_tokens=1024`๋ก ๊ณ ์ ํ์ต๋๋ค.
|
| 155 |
+
- Gemma 4๋ JSON ์ถ๋ ฅ์ ์ํด `enable_thinking=False`๋ฅผ ์ฌ์ฉํ๊ณ , 26B/31B ๊ณ์ด์ ํ๊ฐ ์ฝ๋์์ empty thought channel ์ฒ๋ฆฌ๋ฅผ ์๋ ์ ์ฉํฉ๋๋ค.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 156 |
|
| 157 |
+
## ํ๊ฐ ์ํ
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 158 |
|
| 159 |
+
- Current corrected TB2-lite score: `pending`
|
| 160 |
+
- Reason: ํ์ฌ `/home/work/.data/tb2_lite_eval/corrected_readme_models_vllm` ์ง๊ณ ๊ฒฐ๊ณผ์ ์ด HF repo๋ช
์ด ์ง์ ๋งค์นญ๋์ง ์์์ต๋๋ค.
|
| 161 |
+
- Next step: ๋์ผํ `tb2_lite/scripts/replay_eval.py` ๊ฒฝ๋ก๋ก ํ๊ฐ๋ฅผ ๋๋ฆฐ ๋ค ์ ์ ์นด๋๋ก ์๋ ๊ต์ฒดํฉ๋๋ค.
|
| 162 |
|
| 163 |
+
## ๋ชจ๋ธ๊ตฐ ํด์
|
| 164 |
|
| 165 |
+
- ์ด repo๋ ์์ง ํ์ฌ corrected TB2-lite ์ง๊ณ JSON๊ณผ ์ง์ ๋งค์นญ๋๋ ์ ์๊ฐ ์์ต๋๋ค.
|
| 166 |
+
- TB2-lite ์ ์๋ ์ผ๋ฐ ์ง๋ฅ ๋ฒค์น๋งํฌ๊ฐ ์๋๋ผ ํฐ๋ฏธ๋ next-action JSON ์ฌํ ๋ฅ๋ ฅ์ ์ธก์ ํฉ๋๋ค.
|
| 167 |
+
- ์์ฑ ๋ช
๋ น์ ์ค์ ์คํ ์ ์ sandbox, allowlist, human review ๊ฐ์ ์์ ์ฅ์น๋ฅผ ๊ฑฐ์ณ์ผ ํฉ๋๋ค.
|