salakash
/

SamKash-Tolstoy

Text Generation

Model card Files Files and versions

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

Libraries

How to use salakash/SamKash-Tolstoy with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B")
model = PeftModel.from_pretrained(base_model, "salakash/SamKash-Tolstoy")

How to use salakash/SamKash-Tolstoy with Transformers:

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

pipe = pipeline("text-generation", model="salakash/SamKash-Tolstoy")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("salakash/SamKash-Tolstoy", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

How to use salakash/SamKash-Tolstoy with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/salakash/SamKash-Tolstoy

How to use salakash/SamKash-Tolstoy 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 "salakash/SamKash-Tolstoy" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "salakash/SamKash-Tolstoy",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "salakash/SamKash-Tolstoy" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "salakash/SamKash-Tolstoy",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use salakash/SamKash-Tolstoy with Docker Model Runner:
```
docker model run hf.co/salakash/SamKash-Tolstoy
```

Resources

View closed (4)

Add “compression mode”

#49 opened 4 months ago by

Clean dialogue punctuation and attribution

#48 opened 4 months ago by

Add continuous evaluation + regression tests

#47 opened 4 months ago by

Cross-lingual style alignment with parallel passages

#46 opened 4 months ago by

Reproducibility packaging (big for arXiv credibility)

#45 opened 4 months ago by

Negative sampling: teach what not to do

#44 opened 5 months ago by

Add structural tokens for literature

#43 opened 5 months ago by

Hierarchical prompting: outline → draft → polish

#42 opened 5 months ago by

“Style strength” knob via LoRA scaling

#41 opened 5 months ago by

Adapter routing (Mixture of adapters)

#40 opened 5 months ago by

Add explicit “avoid anachronisms” behavior

#39 opened 5 months ago by

Curriculum learning by length

#38 opened 5 months ago by

DAPT → SFT → DPO (3-step quality upgrade)

#37 opened 5 months ago by

Preference data bootstrapping (cheap but effective)

#36 opened 5 months ago by

Multi author literary adapters + author-mix control

#35 opened 5 months ago by

Add a lightweight style classifier for filtering

#34 opened 5 months ago by

Better segmentation: train on scenes, not entire books

#33 opened 5 months ago by

Add explicit “avoid anachronisms” behavior

#32 opened 5 months ago by

RAG for factual grounding (optional)

#31 opened 5 months ago by

Quantization aware adapter testing

#30 opened 5 months ago by

Add continuous evaluation + regression tests

#29 opened 5 months ago by

Better segmentation: train on scenes, not entire books

#28 opened 5 months ago by

Add a lightweight style classifier for filtering

#27 opened 5 months ago by

Preference data bootstrapping (cheap but effective)

#26 opened 5 months ago by

Multi-author literary adapters + author-mix control

#25 opened 5 months ago by

DAPT → SFT → DPO (3 step quality upgrade)

#24 opened 5 months ago by

Add anti-quote-memorization checks

#23 opened 5 months ago by

Shipping/serving improvements

#22 opened 5 months ago by

Add a “modern-to-classical rewrite” mode

#21 opened 5 months ago by

Create multiple adapters by sub skill

#20 opened 5 months ago by

Add “character + POV” conditioning

#19 opened 5 months ago by

Improve long form coherence with “chapter packing”

#18 opened 5 months ago by

Use a Russian-friendly tokenizer strategy

#17 opened 5 months ago by

Improve dataset hygiene for literary corpora

#16 opened 5 months ago by

Evaluation you can publish in a follow-up paper

#15 opened 5 months ago by

Release two adapters instead of one

#14 opened 5 months ago by

Dual-lingual style transfer (English ↔ Russian)

#13 opened 5 months ago by

Add Russian-language capability (two-stage LoRA)

#12 opened 5 months ago by

Improve dataset hygiene for literary corpora

#11 opened 5 months ago by

🚩 Report: Spam

#9 opened 5 months ago by

change "lora_bias" to 'true'

#7 opened 6 months ago by

Update README.md

#5 opened 6 months ago by

Languages spoken

#2 opened 6 months ago by

Update README.md (Remove Duplicates)

#1 opened 6 months ago by