Instructions to use sylvester-francis/typescript-slm-1.5b-full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sylvester-francis/typescript-slm-1.5b-full with Transformers:

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

pipe = pipeline("text-generation", model="sylvester-francis/typescript-slm-1.5b-full")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("sylvester-francis/typescript-slm-1.5b-full")
model = AutoModelForCausalLM.from_pretrained("sylvester-francis/typescript-slm-1.5b-full")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

llama-cpp-python

How to use sylvester-francis/typescript-slm-1.5b-full with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="sylvester-francis/typescript-slm-1.5b-full",
	filename="gguf/typescript-slm-1.5b-q4_k_m.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use sylvester-francis/typescript-slm-1.5b-full 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 sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf sylvester-francis/typescript-slm-1.5b-full: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 sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf sylvester-francis/typescript-slm-1.5b-full: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 sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

Use Docker

docker model run hf.co/sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

LM Studio
Jan

vLLM

How to use sylvester-francis/typescript-slm-1.5b-full with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sylvester-francis/typescript-slm-1.5b-full"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sylvester-francis/typescript-slm-1.5b-full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

SGLang

How to use sylvester-francis/typescript-slm-1.5b-full 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 "sylvester-francis/typescript-slm-1.5b-full" \
    --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": "sylvester-francis/typescript-slm-1.5b-full",
		"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 "sylvester-francis/typescript-slm-1.5b-full" \
        --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": "sylvester-francis/typescript-slm-1.5b-full",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use sylvester-francis/typescript-slm-1.5b-full with Ollama:
```
ollama run hf.co/sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
```

Unsloth Studio

How to use sylvester-francis/typescript-slm-1.5b-full 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 sylvester-francis/typescript-slm-1.5b-full 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 sylvester-francis/typescript-slm-1.5b-full to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for sylvester-francis/typescript-slm-1.5b-full to start chatting

How to use sylvester-francis/typescript-slm-1.5b-full with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sylvester-francis/typescript-slm-1.5b-full: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": "sylvester-francis/typescript-slm-1.5b-full:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use sylvester-francis/typescript-slm-1.5b-full with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sylvester-francis/typescript-slm-1.5b-full: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 sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use sylvester-francis/typescript-slm-1.5b-full with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf sylvester-francis/typescript-slm-1.5b-full: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 "sylvester-francis/typescript-slm-1.5b-full: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 sylvester-francis/typescript-slm-1.5b-full with Docker Model Runner:
```
docker model run hf.co/sylvester-francis/typescript-slm-1.5b-full:Q4_K_M
```

Lemonade

How to use sylvester-francis/typescript-slm-1.5b-full with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull sylvester-francis/typescript-slm-1.5b-full:Q4_K_M

Run and chat with the model

lemonade run user.typescript-slm-1.5b-full-Q4_K_M

List all available models

lemonade list

TypeScript-SLM-1.5B-Full

TypeScript-SLM-1.5B is a compact, domain-specialized language model fine-tuned for TypeScript code generation, with a focus on React, Next.js, Angular, and Node.js frameworks.

This repository contains the full merged model (base model + LoRA adapters) along with GGUF quantized versions ready for Ollama deployment.

Model Description

Base Model: Qwen/Qwen2.5-Coder-1.5B-Instruct
Model Type: Causal Language Model (Code Generation)
Parameters: 1.5 billion
Context Length: 1024 tokens
Fine-tuning Method: LoRA (Low-Rank Adaptation)
License: MIT
Language: English (Code: TypeScript/JavaScript)

Key Features

✅ Specialized in TypeScript code generation
✅ Framework-aware (React, Next.js, Angular, Node.js)
✅ Strongly-typed code with proper interfaces and types
✅ Optimized for modern web development patterns
✅ Available in multiple GGUF quantizations for Ollama
✅ Fast inference on consumer hardware

Intended Uses

Primary Use Cases

TypeScript Code Completion: Auto-complete TypeScript code in IDEs
Component Generation: Create React/Angular components from descriptions
Type Definition: Generate TypeScript interfaces and type aliases
Code Snippets: Quick generation of framework-specific patterns
Learning Aid: Study TypeScript and framework best practices

Example Prompts

// React component with TypeScript
"Create a React component with TypeScript for a user profile card"

// Next.js API route
"Write a Next.js API route that handles user authentication"

// Angular service
"Create an Angular service for managing todo items"

// Node.js Express server
"Write an Express server with TypeScript that includes CORS and error handling"

// Type definitions
"Define a TypeScript interface for a blog post with author information"

How to Use

Option 1: Ollama (Recommended for Local Use)

The easiest way to use this model locally is with Ollama:

# Import the model using the Modelfile
ollama create typescript-slm-1.5b -f Modelfile-q4_k_m

# Run the model
ollama run typescript-slm-1.5b "Create a React component for a todo list"

Available Quantizations:

Modelfile-q4_k_m - 4-bit quantization (~800MB, fastest)
Modelfile-q6_k - 6-bit quantization (~1.2GB, balanced)
Modelfile-f16 - 16-bit float (~3GB, highest quality)

Option 2: Transformers (Python)

Use directly with the Transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model and tokenizer
model_name = "sylvester-francis/typescript-slm-1.5b-full"
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Generate code
prompt = "Create a React component with TypeScript for a user profile card:"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
    temperature=0.3,
    top_p=0.95,
    do_sample=True,
    pad_token_id=tokenizer.eos_token_id
)

code = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(code)

Option 3: GGUF Files (llama.cpp)

Download GGUF files directly for use with llama.cpp:

# Download specific quantization
huggingface-cli download sylvester-francis/typescript-slm-1.5b-full \
  gguf/typescript-slm-1.5b-q4_k_m.gguf --local-dir ./models

# Run with llama.cpp
./llama-cli -m ./models/gguf/typescript-slm-1.5b-q4_k_m.gguf \
  -p "Create a TypeScript interface for a user profile"

Model Details

Architecture

Base: Qwen2.5-Coder-1.5B-Instruct
Modifications: LoRA fine-tuning on TypeScript-specific data
LoRA Rank: 64
LoRA Alpha: 128
Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Training Data

The model was fine-tuned on a curated dataset of TypeScript code:

Sources: GitHub repositories (1000+ stars), StackOverflow Q&A
Total Samples: ~8,000 high-quality code samples
Quality Filtering: Intelligent scoring based on TypeScript features
Framework Distribution:
- React: ~50% (components, hooks, context)
- Angular: ~25% (services, directives, modules)
- Next.js: ~15% (pages, API routes, SSR)
- Node.js: ~10% (Express, NestJS, APIs)

Quality Indicators:

Proper TypeScript type annotations
Complete modules with imports/exports
Framework-specific best practices
Production-quality code from popular repositories

Training Configuration

Base Model: Qwen/Qwen2.5-Coder-1.5B-Instruct
Training Method: LoRA Fine-tuning
LoRA Rank: 64
LoRA Alpha: 128
Learning Rate: 2e-4
Batch Size: 8 (effective: 32 with gradient accumulation)
Epochs: 3
Max Sequence Length: 1024
Optimizer: AdamW
Warmup Ratio: 0.03

Training Hardware

Platform: Google Colab A100 (40GB)
Training Time: ~30 minutes
Framework: Hugging Face TRL + PEFT

Performance

Code Quality Metrics

Based on evaluation of 100 TypeScript generation tasks:

Metric	Score
Correct Syntax	85%
Proper TypeScript Types	72%
Framework Best Practices	68%
Context Understanding	1024 tokens

Generation Speed

Platform	Tokens/Second
NVIDIA RTX 3090 (FP16)	~80 tokens/s
Apple M2 Max (GGUF q4_k_m)	~45 tokens/s
Apple M1 (GGUF q4_k_m)	~30 tokens/s
CPU (GGUF q4_k_m)	~10-15 tokens/s

Repository Contents

typescript-slm-1.5b-full/
├── config.json                           # Model configuration
├── tokenizer.json                        # Tokenizer configuration
├── tokenizer_config.json                 # Tokenizer settings
├── generation_config.json                # Generation parameters
├── pytorch_model.bin                     # Full merged PyTorch model
├── model.safetensors                     # SafeTensors format
│
└── gguf/                                 # GGUF quantized models
    ├── typescript-slm-1.5b-q4_k_m.gguf  # 4-bit quantization (~800MB)
    ├── typescript-slm-1.5b-q6_k.gguf    # 6-bit quantization (~1.2GB)
    ├── typescript-slm-1.5b-f16.gguf     # 16-bit float (~3GB)
    ├── Modelfile-q4_k_m                  # Ollama config (4-bit)
    ├── Modelfile-q6_k                    # Ollama config (6-bit)
    └── Modelfile-f16                     # Ollama config (16-bit)

Quantization Comparison

Format	Size	Quality	Speed	Use Case
q4_k_m	~800MB	Good	Fastest	Local development, quick iteration
q6_k	~1.2GB	Very Good	Fast	Production, balanced performance
f16	~3GB	Excellent	Moderate	High quality, benchmarking
PyTorch	~3GB	Perfect	GPU-dependent	Fine-tuning, research

Recommendation: Use q4_k_m for testing, q6_k for production.

Limitations

Known Limitations

Context Length: Limited to 1024 tokens (larger contexts may lose coherence)
Complex Logic: May struggle with very complex algorithmic tasks
Newer Frameworks: Limited knowledge of frameworks released after training cutoff
Type Inference: Sometimes requires explicit type annotations
Edge Cases: May not handle all TypeScript edge cases correctly

Not Recommended For

❌ Production-critical code without review
❌ Security-sensitive implementations
❌ Complex algorithm design
❌ Large-scale refactoring
❌ Framework versions beyond training data

Best Used With

✅ Human review and validation
✅ Existing codebase context
✅ Clear, specific prompts
✅ Common framework patterns
✅ Learning and prototyping

Ethical Considerations

Intended Use

This model is designed for:

Developer productivity: Assisting professional developers
Learning: Helping students learn TypeScript and frameworks
Prototyping: Quick generation of boilerplate code

Potential Risks

Code Quality: Generated code should always be reviewed
Security: May generate insecure patterns if prompted
Licensing: Generated code may resemble training data patterns
Bias: May reflect patterns common in open-source code

Responsible Use Guidelines

Always Review: Never deploy generated code without review
Test Thoroughly: All generated code should be tested
Check Licenses: Ensure compliance with relevant licenses
Security Audit: Review for security vulnerabilities
Attribution: Credit the model when appropriate

Related Models

Model Family

typescript-slm-1.5b - LoRA adapter only
typescript-slm-7b - Larger 7B variant
typescript-slm-7b-reasoning - 7B with enhanced reasoning

Comparison

Model	Parameters	Context	Speed	Quality	Use Case
1.5B	1.5B	1024	Fastest	Good	Local dev, quick iteration
7B	7B	2048	Fast	Excellent	Production code
7B-Reasoning	7B	2048	Moderate	Excellent+	Complex problems, debugging

Training Pipeline

This model was created using the TypeScript SLM training pipeline:

# Train your own model
git clone https://github.com/sylvester-francis/slm-typescript-model
cd slm-typescript-model
pip install -r requirements.txt

# Run complete pipeline
python slm.py pipeline

# Deploy to Ollama
python slm.py deploy typescript-slm-1.5b

Citation

If you use this model in your research or project, please cite:

@software{typescript_slm_1.5b_2025,
  author = {Francis, Sylvester},
  title = {TypeScript-SLM-1.5B: Domain-Specialized Language Model for TypeScript},
  year = {2025},
  publisher = {HuggingFace},
  url = {https://huggingface.co/sylvester-francis/typescript-slm-1.5b-full}
}

Acknowledgments

Base Model: Qwen2.5-Coder-1.5B-Instruct by Alibaba Cloud
Training Framework: Hugging Face TRL and PEFT
GGUF Conversion: llama.cpp by Georgi Gerganov

License

This model is released under the MIT License.

MIT License

Copyright (c) 2025 Sylvester Francis

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

Contact & Support

Issues: https://github.com/sylvester-francis/slm-typescript-model/issues
Repository: https://github.com/sylvester-francis/slm-typescript-model
HuggingFace: https://huggingface.co/sylvester-francis

Version History

v1.0.0 (2025-11-29): Initial release
- Full merged model with LoRA adapters
- GGUF quantizations (q4_k_m, q6_k, f16)
- Ollama Modelfiles
- Optimized for React, Next.js, Angular, Node.js

Keywords: typescript, react, nextjs, angular, nodejs, code-generation, llm, gguf, ollama, qwen, lora, small-language-model

Downloads last month: 129

Safetensors

Model size

2B params

Tensor type

BF16

Model tree for sylvester-francis/typescript-slm-1.5b-full

Base model

Qwen/Qwen2.5-1.5B

Finetuned

Qwen/Qwen2.5-Coder-1.5B

Finetuned

Qwen/Qwen2.5-Coder-1.5B-Instruct

Quantized

(141)

this model