Instructions to use Heralax/llama-gRPo-thoughtprocess with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Heralax/llama-gRPo-thoughtprocess with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Heralax/llama-gRPo-thoughtprocess",
	filename="Thoughtbasethoughtdata-Merged-7.2B-Q8_0.gguf",
)

output = llm(
	"Once upon a time,",
	max_tokens=512,
	echo=True
)
print(output)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use Heralax/llama-gRPo-thoughtprocess with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Heralax/llama-gRPo-thoughtprocess:Q8_0
# Run inference directly in the terminal:
llama-cli -hf Heralax/llama-gRPo-thoughtprocess:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Heralax/llama-gRPo-thoughtprocess:Q8_0
# Run inference directly in the terminal:
llama-cli -hf Heralax/llama-gRPo-thoughtprocess:Q8_0

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 Heralax/llama-gRPo-thoughtprocess:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf Heralax/llama-gRPo-thoughtprocess:Q8_0

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 Heralax/llama-gRPo-thoughtprocess:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Heralax/llama-gRPo-thoughtprocess:Q8_0

Use Docker

docker model run hf.co/Heralax/llama-gRPo-thoughtprocess:Q8_0

LM Studio
Jan
Ollama
How to use Heralax/llama-gRPo-thoughtprocess with Ollama:
```
ollama run hf.co/Heralax/llama-gRPo-thoughtprocess:Q8_0
```

Unsloth Studio

How to use Heralax/llama-gRPo-thoughtprocess with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Heralax/llama-gRPo-thoughtprocess to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Heralax/llama-gRPo-thoughtprocess to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Heralax/llama-gRPo-thoughtprocess to start chatting

Docker Model Runner
How to use Heralax/llama-gRPo-thoughtprocess with Docker Model Runner:
```
docker model run hf.co/Heralax/llama-gRPo-thoughtprocess:Q8_0
```

Lemonade

How to use Heralax/llama-gRPo-thoughtprocess with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Heralax/llama-gRPo-thoughtprocess:Q8_0

Run and chat with the model

lemonade run user.llama-gRPo-thoughtprocess-Q8_0

List all available models

lemonade list

Heralax commited on Jun 7, 2025

Commit

94e56a9

verified ·

1 Parent(s): ea171df

Update README.md

Browse files

Files changed (1) hide show

README.md +2 -2

README.md CHANGED Viewed

@@ -18,8 +18,8 @@ Stop sequence == "\*\*Finished.\*\*"
 Related Links:
 - [Augmentoolkit](https://github.com/e-p-armstrong/augmentoolkit)
-- [Augmentoolkit Factual Demo Model (the products of the quickstart)]([TODO](https://huggingface.co/Heralax/llama-Augmentoolkit-Quickstart-Factual-Demo))
-- [gRPo model (no thoughts)](TODO)
 Notes:
 This attempt at getting emotional responses mostly succeeded — when the model writes well, it writes *well*. However an interesting quirk emerged -- the model ended up putting most of its emotional exposition in the thought process, rather than in the final actual visible response. Possibly because it was graded higher when it explained its emotions exhaustively, possibly just because models trained to think with SFT first will pad their responses with thoughts a lot. While I think this overall came out well, it also serves as an example of what to look out for because of this. I think the next version will, in addition to the LLM-as-a-reward function approach, include a function which rewards final answers above a certain length.

 Related Links:
 - [Augmentoolkit](https://github.com/e-p-armstrong/augmentoolkit)
+- [Augmentoolkit Factual Demo Model (the products of the quickstart)](https://huggingface.co/Heralax/llama-Augmentoolkit-Quickstart-Factual-Demo)
+- [gRPo model (no thoughts)](https://huggingface.co/Heralax/llama-gRPo-emotions-nothoughts)
 Notes:
 This attempt at getting emotional responses mostly succeeded — when the model writes well, it writes *well*. However an interesting quirk emerged -- the model ended up putting most of its emotional exposition in the thought process, rather than in the final actual visible response. Possibly because it was graded higher when it explained its emotions exhaustively, possibly just because models trained to think with SFT first will pad their responses with thoughts a lot. While I think this overall came out well, it also serves as an example of what to look out for because of this. I think the next version will, in addition to the LLM-as-a-reward function approach, include a function which rewards final answers above a certain length.