Instructions to use codegeist/codegeist-llm with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use codegeist/codegeist-llm with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-1.7B")
model = PeftModel.from_pretrained(base_model, "codegeist/codegeist-llm")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use codegeist/codegeist-llm 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 codegeist/codegeist-llm:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf codegeist/codegeist-llm:Q4_K_M

Install from WinGet (Windows)

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

Use Docker

docker model run hf.co/codegeist/codegeist-llm:Q4_K_M

LM Studio
Jan

vLLM

How to use codegeist/codegeist-llm with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "codegeist/codegeist-llm"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "codegeist/codegeist-llm",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/codegeist/codegeist-llm:Q4_K_M

Ollama
How to use codegeist/codegeist-llm with Ollama:
```
ollama run hf.co/codegeist/codegeist-llm:Q4_K_M
```

Unsloth Studio

How to use codegeist/codegeist-llm 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 codegeist/codegeist-llm 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 codegeist/codegeist-llm to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for codegeist/codegeist-llm to start chatting

How to use codegeist/codegeist-llm with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf codegeist/codegeist-llm:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "codegeist/codegeist-llm:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

OpenClaw new

How to use codegeist/codegeist-llm with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf codegeist/codegeist-llm:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "codegeist/codegeist-llm:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use codegeist/codegeist-llm with Docker Model Runner:
```
docker model run hf.co/codegeist/codegeist-llm:Q4_K_M
```

Lemonade

How to use codegeist/codegeist-llm with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull codegeist/codegeist-llm:Q4_K_M

Run and chat with the model

lemonade run user.codegeist-llm-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use codegeist/codegeist-llm with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf codegeist/codegeist-llm:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default codegeist/codegeist-llm:Q4_K_M

Run Hermes

hermes

Atomic Chat

codegeist commited on 7 days ago

Commit

312a68f

verified ·

1 Parent(s): 709dcff

Verify strict GPU-only model state

Browse files

Record the final A10G proof that all parameters and buffers are on CUDA, all floating parameters are BF16, and the raw response matches exactly.

Files changed (4) hide show

README.md +8 -5
SHA256SUMS +3 -3
gpu-test-result.json +6 -4
publication.json +28 -2

README.md CHANGED Viewed

@@ -172,22 +172,25 @@ inference repeatability, deterministic PyTorch algorithms, coding benchmarks,
 safety evaluation, and generalization were not tested.
 The successful public-artifact verification ran as Hugging Face Job
-[`6a760e12da2af92a634eedc6`](https://huggingface.co/jobs/codegeist/6a760e12da2af92a634eedc6)
 on NVIDIA A10G. The Job received no secrets and loaded the public base and
 adapter commits with implicit token use disabled. It verified:
 - Adapter weight SHA-256
   `19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8`.
-- CUDA BF16 with every model parameter on the GPU and no CPU fallback.
 - Peak allocated CUDA memory of 3,511,419,904 bytes.
-- A 21.724-second measured load-and-generation phase.
 - Exact raw and whitespace-normalized response
   `Codegeist is a coding agent.`.
 `gpu-test-result.json` contains the sanitized result and source hashes. The Job
-ran for 75 reported seconds. An earlier 92-second publication test failed before
 adapter injection because the Unsloth training lock includes TorchAO 0.13, which
-direct PEFT 0.20 inference rejects. The successful test used a separate locked
 inference environment without Unsloth or TorchAO; the adapter is not
 TorchAO-quantized. CPU inference remains outside the supported contract.

 safety evaluation, and generalization were not tested.
 The successful public-artifact verification ran as Hugging Face Job
+[`6a7610a53e1f34a7e32bd8a8`](https://huggingface.co/jobs/codegeist/6a7610a53e1f34a7e32bd8a8)
 on NVIDIA A10G. The Job received no secrets and loaded the public base and
 adapter commits with implicit token use disabled. It verified:
 - Adapter weight SHA-256
   `19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8`.
+- CUDA BF16 with every floating parameter in BF16 and every parameter and buffer
+  on the GPU, with no CPU fallback.
 - Peak allocated CUDA memory of 3,511,419,904 bytes.
+- A 20.069-second measured load-and-generation phase.
 - Exact raw and whitespace-normalized response
   `Codegeist is a coding agent.`.
 `gpu-test-result.json` contains the sanitized result and source hashes. The Job
+ran for 76 reported seconds. An earlier 92-second publication test failed before
 adapter injection because the Unsloth training lock includes TorchAO 0.13, which
+direct PEFT 0.20 inference rejects. A preliminary 75-second pass then verified
+all parameters on CUDA; the final Job expanded the gate to every buffer and
+every floating-parameter dtype. The successful tests used a separate locked
 inference environment without Unsloth or TorchAO; the adapter is not
 TorchAO-quantized. CPU inference remains outside the supported contract.

SHA256SUMS CHANGED Viewed

@@ -1,8 +1,8 @@
 9a66ed1f77d750a879b0e7b610bb15bb7c109fc1158448c0d7d543e7dbef421f  LICENSE
-c66441fa829c3fd05222fa22d2f6307903bf35ba7652d2df42617367e3a58af0  README.md
 d7ba9293f1820c63fe9e361ab3028390ce646125c898c008a4f8278eed8a4cb5  THIRD_PARTY_NOTICES.md
 42ef1e8075588b3732d0c00dd4c2a08a5e3498429d0e83192210e8006bdf15fb  adapter_config.json
 19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8  adapter_model.safetensors
 502de0994cc784811c28aeb2bb27b478887256489764417543b7493dfd44f7c6  evidence.json
-c5b3e8567fc77050e6074ca944cb5ffca1603b7072d27dca69df9b9c67727939  gpu-test-result.json
-25aaf0ce11772c9466f907c7b1495e981802ef352eac42beedaae42c85ad0c67  publication.json

 9a66ed1f77d750a879b0e7b610bb15bb7c109fc1158448c0d7d543e7dbef421f  LICENSE
+21f2832009be8bb1e8095405467d18fb73ddb1d94eb49230bbfd39108609e5e5  README.md
 d7ba9293f1820c63fe9e361ab3028390ce646125c898c008a4f8278eed8a4cb5  THIRD_PARTY_NOTICES.md
 42ef1e8075588b3732d0c00dd4c2a08a5e3498429d0e83192210e8006bdf15fb  adapter_config.json
 19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8  adapter_model.safetensors
 502de0994cc784811c28aeb2bb27b478887256489764417543b7493dfd44f7c6  evidence.json
+339a15a527229ab82bebce069cb96987a6e2ebb977261f03553759a8f979e57a  gpu-test-result.json
+626d130fd93c5afcca83ef5e3f25cc1012eb1bf7705cc67972c8e08148f3c358  publication.json

gpu-test-result.json CHANGED Viewed

@@ -2,17 +2,19 @@
   "adapter_model": "codegeist/qwen3-1.7b-codegeist-identity-smoke",
   "adapter_revision": "04d51edac56c6f1e068c644bfa8d014cadcecf9f",
   "adapter_weight_sha256": "19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8",
   "all_parameters_on_cuda": true,
   "base_model": "Qwen/Qwen3-1.7B",
   "base_revision": "70d244cc86ccca08cf5af4e1e306ecf908b1ad5e",
   "device": "cuda",
-  "dtype": "bfloat16",
-  "duration_seconds": 21.724,
   "expected_response": "Codegeist is a coding agent.",
   "hardware": "NVIDIA A10G",
   "job": {
     "accelerator": "gpu",
-    "id": "6a760e12da2af92a634eedc6"
   },
   "normalization": "strip leading and trailing whitespace",
   "normalized_match": true,
@@ -32,7 +34,7 @@
     "python": "3.12.12"
   },
   "source_sha256": {
-    "infer.py": "b540a6f584e65fcf5c13811a5133b7477965e5bf6764645cc67620a2728cb50d",
     "inference/pyproject.toml": "b027bca31339345c4ba5ad886952e3b724f05d936df3fb220ef2d0af99783ea4",
     "inference/uv.lock": "ebeda66f1193fbdddd4a06c7e3ac3c7789d78c84c224259246e43214b7031bfa"
   }

   "adapter_model": "codegeist/qwen3-1.7b-codegeist-identity-smoke",
   "adapter_revision": "04d51edac56c6f1e068c644bfa8d014cadcecf9f",
   "adapter_weight_sha256": "19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8",
+  "all_buffers_on_cuda": true,
+  "all_floating_parameters_bfloat16": true,
   "all_parameters_on_cuda": true,
   "base_model": "Qwen/Qwen3-1.7B",
+  "base_model_dtype": "bfloat16",
   "base_revision": "70d244cc86ccca08cf5af4e1e306ecf908b1ad5e",
   "device": "cuda",
+  "duration_seconds": 20.069,
   "expected_response": "Codegeist is a coding agent.",
   "hardware": "NVIDIA A10G",
   "job": {
     "accelerator": "gpu",
+    "id": "6a7610a53e1f34a7e32bd8a8"
   },
   "normalization": "strip leading and trailing whitespace",
   "normalized_match": true,
     "python": "3.12.12"
   },
   "source_sha256": {
+    "infer.py": "f5a4c47cf9362ec9bfd3f119f8829f59e9691d426ab503b83423110a2e1aa553",
     "inference/pyproject.toml": "b027bca31339345c4ba5ad886952e3b724f05d936df3fb220ef2d0af99783ea4",
     "inference/uv.lock": "ebeda66f1193fbdddd4a06c7e3ac3c7789d78c84c224259246e43214b7031bfa"
   }

publication.json CHANGED Viewed

@@ -25,7 +25,7 @@
       "running_seconds": 92,
       "finding": "The Unsloth training lock installs TorchAO 0.13, which direct PEFT 0.20 adapter injection rejects."
     },
-    "successful_job": {
       "id": "6a760e12da2af92a634eedc6",
       "terminal_status": "COMPLETED",
       "running_seconds": 75,
@@ -43,10 +43,36 @@
       "normalized_match": true,
       "result_sha256": "c5b3e8567fc77050e6074ca944cb5ffca1603b7072d27dca69df9b9c67727939"
     },
     "inference_source_sha256": {
-      "infer.py": "b540a6f584e65fcf5c13811a5133b7477965e5bf6764645cc67620a2728cb50d",
       "inference/pyproject.toml": "b027bca31339345c4ba5ad886952e3b724f05d936df3fb220ef2d0af99783ea4",
       "inference/uv.lock": "ebeda66f1193fbdddd4a06c7e3ac3c7789d78c84c224259246e43214b7031bfa"
     }
   },
   "adapter_weights_changed": false,

       "running_seconds": 92,
       "finding": "The Unsloth training lock installs TorchAO 0.13, which direct PEFT 0.20 adapter injection rejects."
     },
+    "preliminary_successful_job": {
       "id": "6a760e12da2af92a634eedc6",
       "terminal_status": "COMPLETED",
       "running_seconds": 75,
       "normalized_match": true,
       "result_sha256": "c5b3e8567fc77050e6074ca944cb5ffca1603b7072d27dca69df9b9c67727939"
     },
+    "successful_job": {
+      "id": "6a7610a53e1f34a7e32bd8a8",
+      "terminal_status": "COMPLETED",
+      "running_seconds": 76,
+      "secrets": [],
+      "hardware": "NVIDIA A10G",
+      "device": "cuda",
+      "base_model_dtype": "bfloat16",
+      "all_floating_parameters_bfloat16": true,
+      "all_parameters_on_cuda": true,
+      "all_buffers_on_cuda": true,
+      "peak_cuda_memory_bytes": 3511419904,
+      "measured_phase_seconds": 20.069,
+      "adapter_revision": "04d51edac56c6f1e068c644bfa8d014cadcecf9f",
+      "adapter_weight_sha256": "19d424106ef88ffeac4c26c22cebfb13ae1d5f309e1dcccf2da708727bec10a8",
+      "raw_response": "Codegeist is a coding agent.",
+      "normalized_response": "Codegeist is a coding agent.",
+      "normalized_match": true,
+      "result_sha256": "339a15a527229ab82bebce069cb96987a6e2ebb977261f03553759a8f979e57a"
+    },
     "inference_source_sha256": {
+      "infer.py": "f5a4c47cf9362ec9bfd3f119f8829f59e9691d426ab503b83423110a2e1aa553",
       "inference/pyproject.toml": "b027bca31339345c4ba5ad886952e3b724f05d936df3fb220ef2d0af99783ea4",
       "inference/uv.lock": "ebeda66f1193fbdddd4a06c7e3ac3c7789d78c84c224259246e43214b7031bfa"
+    },
+    "cost_estimate": {
+      "running_seconds": 243,
+      "per_second_estimate_usd": 0.0675,
+      "conservative_whole_minutes": 6,
+      "conservative_estimate_usd": 0.1002
     }
   },
   "adapter_weights_changed": false,