Instructions to use pathcosmos/frankenstallm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use pathcosmos/frankenstallm with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="pathcosmos/frankenstallm")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("pathcosmos/frankenstallm")
model = AutoModelForCausalLM.from_pretrained("pathcosmos/frankenstallm", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use pathcosmos/frankenstallm with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf pathcosmos/frankenstallm:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pathcosmos/frankenstallm:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf pathcosmos/frankenstallm:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf pathcosmos/frankenstallm:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf pathcosmos/frankenstallm:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf pathcosmos/frankenstallm:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf pathcosmos/frankenstallm:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf pathcosmos/frankenstallm:Q4_K_M

Use Docker

docker model run hf.co/pathcosmos/frankenstallm:Q4_K_M

LM Studio
Jan

vLLM

How to use pathcosmos/frankenstallm with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "pathcosmos/frankenstallm"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pathcosmos/frankenstallm",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/pathcosmos/frankenstallm:Q4_K_M

SGLang

How to use pathcosmos/frankenstallm 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 "pathcosmos/frankenstallm" \
    --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": "pathcosmos/frankenstallm",
		"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 "pathcosmos/frankenstallm" \
        --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": "pathcosmos/frankenstallm",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use pathcosmos/frankenstallm with Ollama:
```
ollama run hf.co/pathcosmos/frankenstallm:Q4_K_M
```
Unsloth Desktop
Docker Model Runner
How to use pathcosmos/frankenstallm with Docker Model Runner:
```
docker model run hf.co/pathcosmos/frankenstallm:Q4_K_M
```

Lemonade

How to use pathcosmos/frankenstallm with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull pathcosmos/frankenstallm:Q4_K_M

Run and chat with the model

lemonade run user.frankenstallm-Q4_K_M

List all available models

lemonade list

Atomic Chat

frankenstallm

Commit History

Mar 26

02:56
Add Related Projects section with EVAFRILL-Mo links

pathcosmos

Verified signature
b9f7777
01:42
docs: 감사의 글/Acknowledgment 한영 분리 (각 섹션 끝에 해당 언어로)

pathcosmos

Verified signature
274410d
01:38
docs: add government acknowledgment (첨단 GPU 활용 지원 사업, MSIT/NIPA)

pathcosmos

Verified signature
adcc67a

Mar 25

07:29
docs: Add file download links for models, GGUF, training data, scripts, and reports

pathcosmos

Verified signature
127e2c7
07:24
fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (Q8_0 3.0GB, byte-fallback)

pathcosmos

Verified signature
25a61c4
07:23
fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (Q4_K_M 1.8GB, byte-fallback)

pathcosmos

Verified signature
c59eb0a
07:20
feat: Add SFT training data (filtered, 2.4M samples from 24 sources)

pathcosmos

Verified signature
6ebad26
07:19
fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (F16 5.7GB, byte-fallback)

pathcosmos

Verified signature
7f2374a
07:15
feat: Add heegyu_orca-math-korean-preference-cleaned.jsonl for ORPO/SFT reproducibility

pathcosmos

Verified signature
449ea1c
07:05
feat: Add nayohan_preference-collection-ko-full.jsonl for ORPO/SFT reproducibility

pathcosmos

Verified signature
b412bf5
06:38
feat: Add combined_preference.jsonl for ORPO/SFT reproducibility

pathcosmos

Verified signature
60d9aab
06:21
feat: Add SFT val + preference data (ORPO training, 630K pairs)

pathcosmos

Verified signature
e9af455
06:06
feat: Add data pipeline scripts + phase reports (Tier 3 - reproducibility)

pathcosmos

Verified signature
b3d361d
06:05
fix: Replace 1.2B model with correct 3B ORPO model (hidden=3072, 28L, vocab=64256, byte-fallback)

pathcosmos

Verified signature
0d9e743
05:42
docs: Add v2 model replacement notice (2026-03-26) - 1.2B→3B fix, update file sizes and vocab

pathcosmos

Verified signature
1e51e84
05:41
fix: Update generation_config.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

pathcosmos

Verified signature
58d3f23
05:41
fix: Update tokenizer_config.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

pathcosmos

Verified signature
8fb78b1
05:41
fix: Update tokenizer.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

pathcosmos

Verified signature
c3a0907
05:41
fix: Replace incorrect 1.2B config with correct 3B ORPO config (hidden_size=3072, 28 layers, vocab=64256)

pathcosmos

Verified signature
e5b6ed2

Mar 10

05:10
docs: add Korean model card + contact info

pathcosmos

Verified signature
db64d78
04:55
docs: add comprehensive HuggingFace model card with benchmarks and usage guide (#51)⁠

pathcosmos +1: somebody-to-love

8b82c63

Mar 9

16:51
docs: add detailed sampling config section with eval grid results

pathcosmos

Verified signature
dc788f2
16:51
docs: add training hardware specs (8x B200, EPYC 9365, software stack)

pathcosmos

Verified signature
ef6210a
16:48
feat: add v1 GGUF gguf/frankenstallm-3b-f16.gguf

pathcosmos

Verified signature
d135462
16:47
feat: add v1 GGUF gguf/frankenstallm-3b-Q8_0.gguf

pathcosmos

Verified signature
a4129b4
16:47
feat: add v1 GGUF gguf/frankenstallm-3b-Q4_K_M.gguf

pathcosmos

Verified signature
0c0a611
16:46
feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-f16)

pathcosmos

Verified signature
0409e36
16:46
feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-Q8_0)

pathcosmos

Verified signature
da759cf
16:46
feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-Q4_K_M)

pathcosmos

Verified signature
b18604b
16:39
feat: add frankenstallm-3b-v2 Q8_0 quantization

pathcosmos

Verified signature
6c8bad0
16:35
feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-Q8_0)

pathcosmos

Verified signature
02abf5c
16:35
feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-f16)

pathcosmos

Verified signature
60b6e86
16:35
feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-Q4_K_M)

pathcosmos

Verified signature
485fd13
16:24
Upload gguf/frankenstallm-3b-v2-Q4_K_M.gguf with huggingface_hub (#50)⁠

pathcosmos +1: somebody-to-love

90181b4
16:24
Upload gguf/frankenstallm-3b-v2-Q4_K_M.gguf with huggingface_hub (#43)⁠

pathcosmos +1: somebody-to-love

0676c94
16:23
Upload gguf/frankenstallm-3b-v2-f16.gguf with huggingface_hub (#42)⁠

pathcosmos +1: somebody-to-love

a9959c7
15:56
Upload gguf/frankenstallm-3b-v2-f16.gguf with huggingface_hub (#49)⁠

pathcosmos +1: somebody-to-love

509cf90
15:51
Upload sampling_config.json with huggingface_hub (#44)⁠

pathcosmos +1: somebody-to-love

dd6eed6
15:51
Upload README.md with huggingface_hub (#45)⁠

pathcosmos +1: somebody-to-love

867657b
15:50
Upload eval/ollama_benchmark_results.json with huggingface_hub (#46)⁠

pathcosmos +1: somebody-to-love

cfa9ba8
15:50
Upload eval_reports/2026-03-09_GGUF_DEPLOYMENT_AND_EVAL_REPORT.md with huggingface_hub (#47)⁠

pathcosmos +1: somebody-to-love

5b02ec5
15:50
Upload folder using huggingface_hub (#41)⁠

pathcosmos +1: somebody-to-love

efd2986
15:50
Upload folder using huggingface_hub (#40)⁠

pathcosmos +1: somebody-to-love

9469fb1
15:49
Upload folder using huggingface_hub (#39)⁠

pathcosmos +1: somebody-to-love

9c3d34f
15:49
Upload eval_reports/2026-03-09_ORPO_EVALUATION_REPORT.md with huggingface_hub (#38)⁠

pathcosmos +1: somebody-to-love

3f46e13
15:49
Upload eval_reports/2026-03-09_GGUF_DEPLOYMENT_AND_EVAL_REPORT.md with huggingface_hub (#37)⁠

pathcosmos +1: somebody-to-love

1d5caa4
15:49
Upload eval/ollama_benchmark_results.json with huggingface_hub (#36)⁠

pathcosmos +1: somebody-to-love

fb7e001
15:49
Upload README.md with huggingface_hub (#35)⁠

pathcosmos +1: somebody-to-love

3d91606
15:49
Upload source/PLAN_hybrid_3b_fixes.md with huggingface_hub (#34)⁠

pathcosmos +1: somebody-to-love

ed8ae31
15:49
Upload source/PROGRESS.md with huggingface_hub (#33)⁠

pathcosmos +1: somebody-to-love

1e78479

Commit History

Mar 26

Add Related Projects section with EVAFRILL-Mo links

docs: 감사의 글/Acknowledgment 한영 분리 (각 섹션 끝에 해당 언어로)

docs: add government acknowledgment (첨단 GPU 활용 지원 사업, MSIT/NIPA)

Mar 25

docs: Add file download links for models, GGUF, training data, scripts, and reports

fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (Q8_0 3.0GB, byte-fallback)

fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (Q4_K_M 1.8GB, byte-fallback)

feat: Add SFT training data (filtered, 2.4M samples from 24 sources)

fix: Replace incorrect 1.2B GGUF with correct 3B ORPO GGUF (F16 5.7GB, byte-fallback)

feat: Add heegyu_orca-math-korean-preference-cleaned.jsonl for ORPO/SFT reproducibility

feat: Add nayohan_preference-collection-ko-full.jsonl for ORPO/SFT reproducibility

feat: Add combined_preference.jsonl for ORPO/SFT reproducibility

feat: Add SFT val + preference data (ORPO training, 630K pairs)

feat: Add data pipeline scripts + phase reports (Tier 3 - reproducibility)

fix: Replace 1.2B model with correct 3B ORPO model (hidden=3072, 28L, vocab=64256, byte-fallback)

docs: Add v2 model replacement notice (2026-03-26) - 1.2B→3B fix, update file sizes and vocab

fix: Update generation_config.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

fix: Update tokenizer_config.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

fix: Update tokenizer.json with correct 3B ORPO byte-fallback tokenizer (vocab=64256)

fix: Replace incorrect 1.2B config with correct 3B ORPO config (hidden_size=3072, 28 layers, vocab=64256)

Mar 10

docs: add Korean model card + contact info

docs: add comprehensive HuggingFace model card with benchmarks and usage guide (#51)⁠

Mar 9

docs: add detailed sampling config section with eval grid results

docs: add training hardware specs (8x B200, EPYC 9365, software stack)

feat: add v1 GGUF gguf/frankenstallm-3b-f16.gguf

feat: add v1 GGUF gguf/frankenstallm-3b-Q8_0.gguf

feat: add v1 GGUF gguf/frankenstallm-3b-Q4_K_M.gguf

feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-f16)

feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-Q8_0)

feat: add v1 Ollama Modelfile (gguf/Modelfile.3b-Q4_K_M)

feat: add frankenstallm-3b-v2 Q8_0 quantization

feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-Q8_0)

feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-f16)

feat: add Ollama Modelfile with sampling config (gguf/Modelfile.3b-v2-Q4_K_M)

Upload gguf/frankenstallm-3b-v2-Q4_K_M.gguf with huggingface_hub (#50)⁠

Upload gguf/frankenstallm-3b-v2-Q4_K_M.gguf with huggingface_hub (#43)⁠

Upload gguf/frankenstallm-3b-v2-f16.gguf with huggingface_hub (#42)⁠

Upload gguf/frankenstallm-3b-v2-f16.gguf with huggingface_hub (#49)⁠

Upload sampling_config.json with huggingface_hub (#44)⁠

Upload README.md with huggingface_hub (#45)⁠

Upload eval/ollama_benchmark_results.json with huggingface_hub (#46)⁠

Upload eval_reports/2026-03-09_GGUF_DEPLOYMENT_AND_EVAL_REPORT.md with huggingface_hub (#47)⁠

Upload folder using huggingface_hub (#41)⁠

Upload folder using huggingface_hub (#40)⁠

Upload folder using huggingface_hub (#39)⁠

Upload eval_reports/2026-03-09_ORPO_EVALUATION_REPORT.md with huggingface_hub (#38)⁠

Upload eval_reports/2026-03-09_GGUF_DEPLOYMENT_AND_EVAL_REPORT.md with huggingface_hub (#37)⁠

Upload eval/ollama_benchmark_results.json with huggingface_hub (#36)⁠

Upload README.md with huggingface_hub (#35)⁠

Upload source/PLAN_hybrid_3b_fixes.md with huggingface_hub (#34)⁠

Upload source/PROGRESS.md with huggingface_hub (#33)⁠