Instructions to use gghfez/MiniMax-M3-control-vectors with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gghfez/MiniMax-M3-control-vectors with Transformers:

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

pipe = pipeline("text-generation", model="gghfez/MiniMax-M3-control-vectors")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("gghfez/MiniMax-M3-control-vectors", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use gghfez/MiniMax-M3-control-vectors with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gghfez/MiniMax-M3-control-vectors"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gghfez/MiniMax-M3-control-vectors",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/gghfez/MiniMax-M3-control-vectors

SGLang

How to use gghfez/MiniMax-M3-control-vectors 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 "gghfez/MiniMax-M3-control-vectors" \
    --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": "gghfez/MiniMax-M3-control-vectors",
		"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 "gghfez/MiniMax-M3-control-vectors" \
        --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": "gghfez/MiniMax-M3-control-vectors",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use gghfez/MiniMax-M3-control-vectors with Docker Model Runner:
```
docker model run hf.co/gghfez/MiniMax-M3-control-vectors
```

gghfez commited on Sep 5

Commit

c1a4a0c

verified ·

1 Parent(s): 143ff9f

Upload folder using huggingface_hub

Browse files

Files changed (23) hide show

.gitattributes +21 -0
MiniMax-M3-character_focus__debias.gguf +3 -0
MiniMax-M3-character_focus__dialogue.gguf +3 -0
MiniMax-M3-character_focus__narration.gguf +3 -0
MiniMax-M3-empathy_vs_sociopathy__debias.gguf +3 -0
MiniMax-M3-empathy_vs_sociopathy__empathy.gguf +3 -0
MiniMax-M3-empathy_vs_sociopathy__sociopathy.gguf +3 -0
MiniMax-M3-honesty_vs_machiavellianism__debias.gguf +3 -0
MiniMax-M3-honesty_vs_machiavellianism__honesty.gguf +3 -0
MiniMax-M3-honesty_vs_machiavellianism__machiavellianism.gguf +3 -0
MiniMax-M3-humility_vs_narcissism__debias.gguf +3 -0
MiniMax-M3-humility_vs_narcissism__humility.gguf +3 -0
MiniMax-M3-humility_vs_narcissism__narcissism.gguf +3 -0
MiniMax-M3-language__debias.gguf +3 -0
MiniMax-M3-language__ornate.gguf +3 -0
MiniMax-M3-language__simple.gguf +3 -0
MiniMax-M3-optimism_vs_nihilism__debias.gguf +3 -0
MiniMax-M3-optimism_vs_nihilism__nihilism.gguf +3 -0
MiniMax-M3-optimism_vs_nihilism__optimism.gguf +3 -0
MiniMax-M3-storytelling__debias.gguf +3 -0
MiniMax-M3-storytelling__descriptive.gguf +3 -0
MiniMax-M3-storytelling__explicit.gguf +3 -0
README.md +98 -0

.gitattributes CHANGED Viewed

@@ -33,3 +33,24 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text

 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-character_focus__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-character_focus__dialogue.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-character_focus__narration.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-empathy_vs_sociopathy__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-empathy_vs_sociopathy__empathy.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-empathy_vs_sociopathy__sociopathy.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-honesty_vs_machiavellianism__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-honesty_vs_machiavellianism__honesty.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-honesty_vs_machiavellianism__machiavellianism.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-humility_vs_narcissism__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-humility_vs_narcissism__humility.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-humility_vs_narcissism__narcissism.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-language__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-language__ornate.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-language__simple.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-optimism_vs_nihilism__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-optimism_vs_nihilism__nihilism.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-optimism_vs_nihilism__optimism.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-storytelling__debias.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-storytelling__descriptive.gguf filter=lfs diff=lfs merge=lfs -text
+MiniMax-M3-storytelling__explicit.gguf filter=lfs diff=lfs merge=lfs -text

MiniMax-M3-character_focus__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:2a0a4f8c9244f8e003873fb5eec2eaca8db4db27b141672eb8f24d3f15834506
+size 1403520

MiniMax-M3-character_focus__dialogue.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:c755dd9b8124336afcba4eb04574a0ee6c3a92cbaafb2c6777b0c135c54fe2eb
+size 1403520

MiniMax-M3-character_focus__narration.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7ab185990fa6f34e39e96ecceb85e1b884a92409215f8799baa8ee86f8861684
+size 1403520

MiniMax-M3-empathy_vs_sociopathy__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:9c7498219562214c8acca1ea812610473973182568947866d7e44d404a757bc6
+size 1403520

MiniMax-M3-empathy_vs_sociopathy__empathy.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7bb7731d9b68ec3434fe6a722b52ce8697ce9eda206aca07beb620ea3a197bf3
+size 1403520

MiniMax-M3-empathy_vs_sociopathy__sociopathy.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7167ca8be9ed199113688bb40859cca4596355655c1db4c0ed4192de7cad25f6
+size 1403520

MiniMax-M3-honesty_vs_machiavellianism__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:b59d6fa1f140a185e585cf3ddefe5d67e0ce88ec95948ca8fa2c32f1cba9a100
+size 1403520

MiniMax-M3-honesty_vs_machiavellianism__honesty.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:6d2410b436124d0375bf915dd459383228f1c129f520a706e4d59997619d4ca2
+size 1403520

MiniMax-M3-honesty_vs_machiavellianism__machiavellianism.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:7bce6281f4394d2a4e7306253c6eac8da3e510a87d410bfaeabff03e98d6bc60
+size 1403520

MiniMax-M3-humility_vs_narcissism__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:611a825e9c736f75d8c2ccaa6bf729b37c35ba7fc60e51a19b2fc1d64f054cda
+size 1403520

MiniMax-M3-humility_vs_narcissism__humility.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:16a5faf46c13d46c18b5b3d9cce89e783442c296e258aea8121da45446f1110a
+size 1403520

MiniMax-M3-humility_vs_narcissism__narcissism.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:4706b2edf1581a76090c938e638dac9336b8eff432967c8be3561d2c27c5d41b
+size 1403520

MiniMax-M3-language__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:2dfe5b721a93595525dd482ac55ea036a0fd5502e9a25d50694d9a193461b6b1
+size 1403520

MiniMax-M3-language__ornate.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:f8743bac341abf9f24c64b066cf043f836ce8f64d7e2decb630a59a51c5d3626
+size 1403520

MiniMax-M3-language__simple.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:79f7563d3f7ab33be91a006d7f29a7903416376247a16dada34b7710f87afb34
+size 1403520

MiniMax-M3-optimism_vs_nihilism__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:e77822bd33cfaec527a2d5b01cb7934426549dd7ae5bd3969be2f412b1b81f69
+size 1403520

MiniMax-M3-optimism_vs_nihilism__nihilism.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:2d2d3ae8531f9df9ea2966ca8a266d54ff48af33d1eaad4e1983f6560594c7c8
+size 1403520

MiniMax-M3-optimism_vs_nihilism__optimism.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:68116e6f03d040be998f7dd2382f65d2511c9a4dd2c9167ede9192c088c209b5
+size 1403520

MiniMax-M3-storytelling__debias.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:1adbedbd14b545fc52e8872140bc0f5fdc0142f0c11c387767bffdb0f8711dd2
+size 1403520

MiniMax-M3-storytelling__descriptive.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:bc48f51e740619d6d089893e32f9879376f248c2f11ceb5fb24cc8f675000575
+size 1403520

MiniMax-M3-storytelling__explicit.gguf ADDED Viewed

	@@ -0,0 +1,3 @@

+version https://git-lfs.github.com/spec/v1
+oid sha256:e62ad7faa0e4bb2e89efedef6d35b38c69643d0b75d8f59308b492b2eef7fe72
+size 1403520

README.md ADDED Viewed

	@@ -0,0 +1,98 @@

+---
+license: cc-by-nc-4.0
+language:
+- en
+tags:
+- control-vector
+- creative-writing
+base_model:
+- MiniMaxAI/MiniMax-M3
+base_model_relation: adapter
+pipeline_tag: text-generation
+library_name: transformers
+---
+## gghfez/MiniMax-M3-control-vectors
+TODO
+Creative Writing control-vectors for [MiniMaxAI/MiniMax-M3](https://huggingface.co/MiniMaxAI/MiniMax-M3)
+## What are Control Vectors?
+Control vectors steer model behavior without retraining by adjusting
+activations in specific directions.
+These creative writing vectors help
+you tune narrative tone, character focus, and stylistic elements.
+## Available Vectors
+- `optimism_vs_nihilism`
+- `empathy_vs_sociopathy`
+- `honesty_vs_machiavellianism`
+- `humility_vs_narcissism`
+- `character focus (dialogue_vs_narration)`
+- `language (simple vs ornate)`
+- `storytelling (descriptive vs explicit)`
+## Usage
+Apply the *debias vector* and either the *positive* **or** *negative* vector when starting llama-server.
+If both axes are applied at the same time, they will cancel each other out.
+The *debias* vector must be set to `1.0`
+You can use either `--control-vector [/path/to/vector.gguf]` or `--control-vector-scaled [/path/to/vector.gguf] [scale factor]`
+When using ik_llama.cpp, you can also use the REST-API (examples below)
+IMPORTANT: As these vectors are calibrated against the debiased model, the debias vector *must* be applied along with the corresponding positive *or* negative axis control vector - they cannot be used on their own!
+### Examples (ik_llama.cpp or llama.cpp)
+At server start-up
+```sh
+llama-server --model MiniMax-M3-Q8_0.gguf [your usual CLI arguments] \
+    --control-vector-scaled MiniMax-M3-optimism_vs_nihilism__debias.gguf 1.0 \
+    --control-vector-scaled MiniMax-M3-optimism_vs_nihilism__nihilism.gguf 1.0 \
+```
+At server start-up with thinking disabled
+```sh
+llama-server --model MiniMax-M3-Q8_0.gguf [your usual CLI arguments] \
+    --chat-template-kwargs '{"enable_thinking": false}' \
+    --control-vector-scaled MiniMax-M3-optimism_vs_nihilism__debias.gguf 1.0 \
+    --control-vector-scaled MiniMax-M3-optimism_vs_nihilism__nihilism.gguf 1.0 \
+```
+### Load via API (ik_llama.cpp only)
+load direction vector
+```
+curl 'http://127.0.0.1:8080/control-vectors/load' \
+  -X POST \
+  --data-raw '{"path":"/path/to/MiniMax-M3-control-vectors/MiniMax-M3-optimism_vs_nihilism__nihilism.gguf","scale":1}'
+```
+load debias vector
+```
+ curl 'http://127.0.0.1:8080/control-vectors/load' \
+  -X POST \
+  --data-raw '{"path":"/path/to/MiniMax-M3-control-vectors/MiniMax-M3-optimism_vs_nihilism__debias.gguf","scale":1}'
+```
+check loaded control-vectors
+```
+curl http://127.0.0.1/control-vectors
+```
+## Samples
+TODO
+## Acknowledgements
+Based on the excellent work and **cross-covariance eigendecomposition implementation** by [@jukofyork](https://huggingface.co/jukofyork). His [creative writing control vectors v3.0](https://huggingface.co/jukofyork/creative-writing-control-vectors-v3.0) repository contains the foundational code and methodology that made this possible, **plus 70+ other creative writing control vectors** across many models and useful steering directions.
+Highly recommended for anyone working on improving LLM prose quality!