Instructions to use rain1955/emotion-vector-replication with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rain1955/emotion-vector-replication with Transformers:

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

pipe = pipeline("text-generation", model="rain1955/emotion-vector-replication")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("rain1955/emotion-vector-replication", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use rain1955/emotion-vector-replication with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rain1955/emotion-vector-replication"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rain1955/emotion-vector-replication",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/rain1955/emotion-vector-replication

SGLang

How to use rain1955/emotion-vector-replication 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 "rain1955/emotion-vector-replication" \
    --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": "rain1955/emotion-vector-replication",
		"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 "rain1955/emotion-vector-replication" \
        --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": "rain1955/emotion-vector-replication",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use rain1955/emotion-vector-replication with Docker Model Runner:
```
docker model run hf.co/rain1955/emotion-vector-replication
```

rain1955 commited on Apr 6

Commit

8e7c29e

verified ·

1 Parent(s): feb586c

Fix arousal groupings per @dejanseo feedback: revised emotion categories, added threshold check for unclear axes

Browse files

Files changed (1) hide show

analyze_vectors.py +10 -4

analyze_vectors.py CHANGED Viewed

@@ -77,8 +77,12 @@ def valence_arousal_check(emotions, pca_results):
     positive = {"happy", "proud", "inspired", "loving", "hopeful", "calm", "playful"}
     negative = {"sad", "angry", "afraid", "desperate", "guilty", "disgusted", "lonely", "spiteful"}
-    high_arousal = {"angry", "afraid", "surprised", "desperate", "inspired", "nervous", "anxious"}
-    low_arousal = {"calm", "sad", "brooding", "lonely", "guilty"}
     for pc_name, pc_vals in [("PC1", pca_results["pc1"]), ("PC2", pca_results["pc2"])]:
         pos_vals = [pc_vals[i] for i, e in enumerate(pca_results["emotions"]) if e in positive]
@@ -98,10 +102,12 @@ def valence_arousal_check(emotions, pca_results):
         print(f"    Positive mean: {pos_mean:+.3f}  Negative mean: {neg_mean:+.3f}  → Valence separation: {valence_sep:.3f}")
         print(f"    High-A mean:   {hi_mean:+.3f}  Low-A mean:    {lo_mean:+.3f}  → Arousal separation: {arousal_sep:.3f}")
-        if valence_sep > arousal_sep:
             print(f"    → {pc_name} ≈ VALENCE axis")
-        else:
             print(f"    → {pc_name} ≈ AROUSAL axis")
 def main():

     positive = {"happy", "proud", "inspired", "loving", "hopeful", "calm", "playful"}
     negative = {"sad", "angry", "afraid", "desperate", "guilty", "disgusted", "lonely", "spiteful"}
+    # Arousal groupings revised per dejanseo's feedback (2026-04-06):
+    # - Removed inspired from high arousal (more contemplative/medium)
+    # - Added disgusted, confused, playful, spiteful to high arousal
+    # - Added loving, hopeful to low arousal
+    high_arousal = {"angry", "afraid", "surprised", "desperate", "nervous", "anxious", "disgusted", "confused", "playful", "spiteful"}
+    low_arousal = {"calm", "sad", "brooding", "lonely", "guilty", "loving", "hopeful"}
     for pc_name, pc_vals in [("PC1", pca_results["pc1"]), ("PC2", pca_results["pc2"])]:
         pos_vals = [pc_vals[i] for i, e in enumerate(pca_results["emotions"]) if e in positive]
         print(f"    Positive mean: {pos_mean:+.3f}  Negative mean: {neg_mean:+.3f}  → Valence separation: {valence_sep:.3f}")
         print(f"    High-A mean:   {hi_mean:+.3f}  Low-A mean:    {lo_mean:+.3f}  → Arousal separation: {arousal_sep:.3f}")
+        if valence_sep > arousal_sep and valence_sep > 1.0:
             print(f"    → {pc_name} ≈ VALENCE axis")
+        elif arousal_sep > valence_sep and arousal_sep > 0.5:
             print(f"    → {pc_name} ≈ AROUSAL axis")
+        else:
+            print(f"    → {pc_name} ≈ UNCLEAR — neither valence nor arousal dominant (model may have learned its own geometry)")
 def main():