Text Generation
Transformers
GGUF
English
Hebrew
gemma4
image-text-to-text
code
python
typescript
coding-assistant
llama.cpp
ollama
unsloth
qlora
on-device
private-first
conversational
Instructions to use BrainboxAI/code-il-E4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use BrainboxAI/code-il-E4B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="BrainboxAI/code-il-E4B")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("BrainboxAI/code-il-E4B") model = AutoModelForMultimodalLM.from_pretrained("BrainboxAI/code-il-E4B", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use BrainboxAI/code-il-E4B 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 BrainboxAI/code-il-E4B:BF16 # Run inference directly in the terminal: llama cli -hf BrainboxAI/code-il-E4B:BF16
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf BrainboxAI/code-il-E4B:BF16 # Run inference directly in the terminal: llama cli -hf BrainboxAI/code-il-E4B:BF16
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 BrainboxAI/code-il-E4B:BF16 # Run inference directly in the terminal: ./llama-cli -hf BrainboxAI/code-il-E4B:BF16
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 BrainboxAI/code-il-E4B:BF16 # Run inference directly in the terminal: ./build/bin/llama-cli -hf BrainboxAI/code-il-E4B:BF16
Use Docker
docker model run hf.co/BrainboxAI/code-il-E4B:BF16
- LM Studio
- Jan
- vLLM
How to use BrainboxAI/code-il-E4B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "BrainboxAI/code-il-E4B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "BrainboxAI/code-il-E4B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/BrainboxAI/code-il-E4B:BF16
- SGLang
How to use BrainboxAI/code-il-E4B 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 "BrainboxAI/code-il-E4B" \ --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": "BrainboxAI/code-il-E4B", "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 "BrainboxAI/code-il-E4B" \ --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": "BrainboxAI/code-il-E4B", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Ollama
How to use BrainboxAI/code-il-E4B with Ollama:
ollama run hf.co/BrainboxAI/code-il-E4B:BF16
- Unsloth Desktop
- Pi
How to use BrainboxAI/code-il-E4B with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf BrainboxAI/code-il-E4B:BF16
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "BrainboxAI/code-il-E4B:BF16" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use BrainboxAI/code-il-E4B with Docker Model Runner:
docker model run hf.co/BrainboxAI/code-il-E4B:BF16
- Lemonade
How to use BrainboxAI/code-il-E4B with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull BrainboxAI/code-il-E4B:BF16
Run and chat with the model
lemonade run user.code-il-E4B-BF16
List all available models
lemonade list
- Hermes Agent
How to use BrainboxAI/code-il-E4B with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf BrainboxAI/code-il-E4B:BF16
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 BrainboxAI/code-il-E4B:BF16
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use BrainboxAI/code-il-E4B with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf BrainboxAI/code-il-E4B:BF16
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 "BrainboxAI/code-il-E4B:BF16" \ --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"
Rewrite the model card in the BrainboxAI house style (Hebrew, brand name in the heading, repository id unchanged)
Browse files
README.md
CHANGED
|
@@ -24,58 +24,70 @@ tags:
|
|
| 24 |
- text-generation
|
| 25 |
- on-device
|
| 26 |
- private-first
|
| 27 |
-
pretty_name:
|
| 28 |
model-index:
|
| 29 |
- name: code-il-E4B
|
| 30 |
results: []
|
| 31 |
---
|
| 32 |
|
| 33 |
-
#
|
| 34 |
|
| 35 |
-
|
|
|
|
|
|
|
| 36 |
|
| 37 |
[](https://huggingface.co/BrainboxAI/code-il-E4B)
|
| 38 |
[](https://huggingface.co/datasets/BrainboxAI/code-training-il)
|
| 39 |
[](https://huggingface.co/BrainboxAI/code-il-E4B-safetensors)
|
| 40 |
[](https://www.apache.org/licenses/LICENSE-2.0)
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
---
|
| 43 |
|
| 44 |
-
##
|
|
|
|
|
|
|
| 45 |
|
| 46 |
-
|
| 47 |
|
| 48 |
-
|
| 49 |
-
-
|
| 50 |
-
-
|
| 51 |
-
- Apple Silicon via llama.cpp Metal
|
| 52 |
|
| 53 |
-
|
| 54 |
|
| 55 |
-
##
|
| 56 |
|
| 57 |
-
|
| 58 |
|
| 59 |
-
|
| 60 |
|
| 61 |
-
|
| 62 |
|
| 63 |
-
##
|
| 64 |
|
| 65 |
-
|
| 66 |
-
-
|
| 67 |
-
-
|
| 68 |
-
-
|
| 69 |
-
-
|
| 70 |
-
- Hebrew-speaking developer onboarding (model responds in Hebrew on request)
|
| 71 |
|
| 72 |
-
**
|
| 73 |
-
- Replacement for frontier models on complex architecture tasks
|
| 74 |
-
- Production code generation without human review
|
| 75 |
-
- Languages other than Python / TypeScript (coverage is minimal)
|
| 76 |
-
- Fine-tuning tasks requiring >4B parameters of capacity
|
| 77 |
|
| 78 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
|
| 80 |
### Ollama
|
| 81 |
|
|
@@ -86,13 +98,15 @@ ollama run hf.co/BrainboxAI/code-il-E4B:Q4_K_M
|
|
| 86 |
|
| 87 |
### llama.cpp
|
| 88 |
|
|
|
|
|
|
|
| 89 |
```bash
|
| 90 |
-
./llama-cli -m
|
| 91 |
-p "Write a Python function that parses ISO-8601 dates with timezones." \
|
| 92 |
--temp 0.2 --top-p 0.95 -n 1024
|
| 93 |
```
|
| 94 |
|
| 95 |
-
###
|
| 96 |
|
| 97 |
```python
|
| 98 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
@@ -112,31 +126,24 @@ outputs = model.generate(inputs, max_new_tokens=1024, temperature=0.2, top_p=0.9
|
|
| 112 |
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 113 |
```
|
| 114 |
|
| 115 |
-
###
|
| 116 |
-
|
| 117 |
-
| Parameter | Value | Rationale |
|
| 118 |
-
|-----------|-------|-----------|
|
| 119 |
-
| `temperature` | 0.2 | Low creativity for deterministic code |
|
| 120 |
-
| `top_p` | 0.95 | Slightly higher than legal model to allow idiom variety |
|
| 121 |
-
| `max_new_tokens` | 1024 | Enough for most function-level completions |
|
| 122 |
-
| `repetition_penalty` | 1.0 | Penalizing repetition hurts code structure |
|
| 123 |
-
|
| 124 |
|
| 125 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
-
|
| 128 |
|
| 129 |
-
|
| 130 |
|
| 131 |
-
|
| 132 |
|
| 133 |
-
|
| 134 |
-
2. **Edge Cases and Constraints** - enumerate what could go wrong before coding
|
| 135 |
-
3. **Implementation** - the actual code, with inline comments only where needed
|
| 136 |
-
4. **Tests** - concrete test cases covering happy path + edge cases
|
| 137 |
-
5. **Known Limitations** - what this code does NOT handle, dependencies, assumptions
|
| 138 |
|
| 139 |
-
###
|
| 140 |
|
| 141 |
```text
|
| 142 |
DEFINITIONS:
|
|
@@ -215,7 +222,7 @@ VERIFICATION:
|
|
| 215 |
- regression check: No "production-ready" claims unless edge cases match limitations.
|
| 216 |
```
|
| 217 |
|
| 218 |
-
###
|
| 219 |
|
| 220 |
```python
|
| 221 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
@@ -227,7 +234,7 @@ model = AutoModelForCausalLM.from_pretrained(
|
|
| 227 |
device_map="auto",
|
| 228 |
)
|
| 229 |
|
| 230 |
-
#
|
| 231 |
SYSTEM_PROMPT = """[paste the full prompt from the code block above]"""
|
| 232 |
|
| 233 |
messages = [
|
|
@@ -240,72 +247,90 @@ outputs = model.generate(inputs, max_new_tokens=1500, temperature=0.2, top_p=0.9
|
|
| 240 |
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 241 |
```
|
| 242 |
|
| 243 |
-
###
|
| 244 |
|
| 245 |
-
-
|
| 246 |
-
-
|
| 247 |
-
-
|
| 248 |
-
-
|
| 249 |
|
|
|
|
| 250 |
|
| 251 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 252 |
|
| 253 |
-
|
| 254 |
-
|-----------|-------|
|
| 255 |
-
| **Base model** | [unsloth/gemma-4-E4B-it](https://huggingface.co/unsloth/gemma-4-E4B-it) |
|
| 256 |
-
| **Method** | QLoRA (4-bit quantization during training) |
|
| 257 |
-
| **LoRA rank (r)** | 64 |
|
| 258 |
-
| **LoRA alpha** | 128 |
|
| 259 |
-
| **Training data size** | 40,000 curated examples |
|
| 260 |
-
| **Train / validation split** | 95% / 5%, seed 3407 |
|
| 261 |
-
| **Hardware** | NVIDIA RTX 5090 (RunPod) |
|
| 262 |
-
| **Framework** | Unsloth Studio |
|
| 263 |
|
| 264 |
-
###
|
| 265 |
|
| 266 |
-
|
|
| 267 |
-
|---
|
| 268 |
-
| [OpenCodeInstruct
|
| 269 |
-
| [typescript-instruct
|
| 270 |
-
|
|
|
|
|
| 271 |
|
| 272 |
-
|
| 273 |
|
| 274 |
-
|
| 275 |
|
| 276 |
-
|
| 277 |
|
| 278 |
-
|
| 279 |
|
| 280 |
-
|
| 281 |
-
|------|----------|--------|-------|
|
| 282 |
-
| **FizzBuzz** (via agentic loop) | 5 | 5/5 | Solved in 6 steps, zero correction rounds |
|
| 283 |
-
| **Binary search with 11 edge cases** | 11 | 11/11 | Including leftmost-duplicate handling |
|
| 284 |
|
| 285 |
-
|
| 286 |
|
| 287 |
-
|
|
|
|
|
|
|
|
|
|
| 288 |
|
| 289 |
-
|
| 290 |
-
- **Two languages.** Strong on Python and TypeScript; weak on other languages.
|
| 291 |
-
- **No tool use out of the box.** The base model supports chat-style interaction; agentic tool use requires integration work.
|
| 292 |
-
- **Training cutoff.** Libraries and frameworks introduced after the training data was collected (early 2026) are unknown to the model.
|
| 293 |
-
- **Hallucination risk.** Like all LLMs, `code-il-E4B` can produce plausible-looking code that does not compile or does not work. Always test.
|
| 294 |
|
| 295 |
-
|
| 296 |
|
| 297 |
-
|
| 298 |
-
- [**Safetensors 16-bit**](https://huggingface.co/BrainboxAI/code-il-E4B-safetensors) โ for further fine-tuning, HF transformers
|
| 299 |
|
| 300 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 301 |
|
| 302 |
-
|
| 303 |
|
| 304 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 305 |
|
| 306 |
```bibtex
|
| 307 |
@misc{elyasi2026codeil,
|
| 308 |
-
title = {Code-IL E4B: A Small, On-Device Coding Assistant for Private Environments},
|
| 309 |
author = {Elyasi, Netanel},
|
| 310 |
year = {2026},
|
| 311 |
publisher = {BrainboxAI},
|
|
@@ -314,12 +339,10 @@ Apache 2.0. Use commercially, modify, and redistribute with attribution.
|
|
| 314 |
}
|
| 315 |
```
|
| 316 |
|
| 317 |
-
##
|
| 318 |
|
| 319 |
-
|
| 320 |
|
| 321 |
-
|
| 322 |
-
|
| 323 |
-
---
|
| 324 |
|
| 325 |
-
*
|
|
|
|
| 24 |
- text-generation
|
| 25 |
- on-device
|
| 26 |
- private-first
|
| 27 |
+
pretty_name: bx-code-nogah (code-il-E4B)
|
| 28 |
model-index:
|
| 29 |
- name: code-il-E4B
|
| 30 |
results: []
|
| 31 |
---
|
| 32 |
|
| 33 |
+
# bx-code-nogah
|
| 34 |
|
| 35 |
+
### ืืืื ืืืืืจ: `BrainboxAI/code-il-E4B`
|
| 36 |
+
|
| 37 |
+
**ืขืืืจ ืืชืืืช ืงืื ื-Python ืื-TypeScript ืฉืจืฅ ืืืื ืขื ืืืืฉื ืฉืื. ืฉืืจืช ืงืื ืื ืืืฆืืช ืืืืฆื.**
|
| 38 |
|
| 39 |
[](https://huggingface.co/BrainboxAI/code-il-E4B)
|
| 40 |
[](https://huggingface.co/datasets/BrainboxAI/code-training-il)
|
| 41 |
[](https://huggingface.co/BrainboxAI/code-il-E4B-safetensors)
|
| 42 |
[](https://www.apache.org/licenses/LICENSE-2.0)
|
| 43 |
|
| 44 |
+
> **ืขื ืืฉื.** `bx-code-nogah` ืืื ืฉื ืืืืื ืืืืกืืืช ืืฉืืืช ืฉื BrainboxAI: `bx` ืืืืจื, `code` ืืชืืื, ื-`nogah` (ื ืืื) ืืืจืืช ืืืืื ืืืืฆืขืืช. **ืืืื ืืืืืจ ื ืฉืืจ `BrainboxAI/code-il-E4B` ืืื ืืฉืชื ื** โ ืื ืงืืฉืืจ ืืกืงืจืืคื ืงืืื ืืืฉืื ืืขืืื.
|
| 45 |
+
|
| 46 |
+
> **ืขื ืืฆืืืืช ืืืจืกื.** ืืืืื ืืืฉ ืขื ืืืชื ืืฉืืื ื ืืืฃ ืืืืชื ืืืืจ ืืืขืืื ืืช ืืืฉืงืืืืช ืืืงืื. ืืืืืจ ืื ืฉืืืจืื ืืืื ืืฉืื ืืขืื ืืืืฉืืื ืขืืื ืืงืื ืืฉืงืืืืช ืฉืื ืืช ืชืืช ืืืชื ืฉื. ืื ืฉืฆืจืื ืืฆืืืืช ืืืืืืช โ ืืฆืืื ืืช ืขืฆืื ื-commit ืืกืืื ืืื ืืขื ืฃ ืืจืืฉื.
|
| 47 |
+
|
| 48 |
+
> **English.** `code-il-E4B` (brand name `bx-code-nogah`) is an on-device Python and TypeScript coding assistant, fine-tuned from `unsloth/gemma-4-E4B-it` on a 40,330-example test-filtered corpus. No formal benchmark (HumanEval, MBPP) has been run on it. This card is in Hebrew; identifiers, code and the recommended system prompt are in English.
|
| 49 |
+
|
| 50 |
---
|
| 51 |
|
| 52 |
+
## ืื ืื, ืืงืฆืจื
|
| 53 |
+
|
| 54 |
+
ืืืื ืฉืืืชื ืืืืืง ืงืื ื-Python ืื-TypeScript, ืืจืฅ ืืฆืื ืขื ืืืืฉื. ืืื ืื ืื ืขื [`unsloth/gemma-4-E4B-it`](https://huggingface.co/unsloth/gemma-4-E4B-it) ืฉื ืืืื, ืืืืื ืืขืืื ืขื 40,330 ืืืืืืืช ืฉืกืื ื ื ืืคื ืงืจืืืจืืื ืืื ืคืฉืื: **ืืื ืืงืื ืืืืืื ืืืืช ืขืืจ ืืช ืืืืืงืืช ืฉืื.**
|
| 55 |
|
| 56 |
+
ืืื ื ืื ืก ืืงืืืฅ ืืื ืฉื ื-5.3 ื'ืืื-ืืืื. ืืื ืจืฅ ืขื:
|
| 57 |
|
| 58 |
+
- ืืขืื ืฉื ืืืฉื ื ืืื ืืืืจื ื โ ืืืื, ืืื ืขืืื.
|
| 59 |
+
- ืื ืืจืืืก ืืกื ืืืชื ืขื 6 ื'ืืื ืืืืจืื ืืืขืื.
|
| 60 |
+
- ืืืฉืื Apple ืขื ืฉืื M, ืืจื llama.cpp.
|
|
|
|
| 61 |
|
| 62 |
+
ืืื ืืืืืจ ืืืื ืืจื ื, ืืื ืืืืื, ืืื ืฉืฉืืจืช ืงืื ืืืช ืืืฆืืช ืืืืืื ื.
|
| 63 |
|
| 64 |
+
## ืืื ืืื ืงืืื
|
| 65 |
|
| 66 |
+
ืื ืชื ืฉื ืฉืื ืืขืืืจ ืงืื ืืขื ื ืืื ืืืืคื ืคืืื ืฆืืืืืช. ืืืืจื ืฉืืื ื ืืขืจืืช ืงื ืืื ืืช โ ืืืืืืื ืืคืื ื ืกืื, ืืจืคืืื ืืืืืืืื โ ืื ืคืฉืื ืื ืขืืืจ.
|
| 67 |
|
| 68 |
+
ืืืืื ืืื ืืื ืืืืืคื ืืคืจืืืช: ืงืื ืืกืคืืง ืืจืืฅ ืืงืืืืช, ืืืืื ืืฉืชื ืืฉืคืืช ืฉืจืื ืืืืจืืช ืืืืช ืืืชืืืช ืืื.
|
| 69 |
|
| 70 |
+
**ืืื ืื ืืชืืจื ื-Claude ืื ื-GPT ืืืืืืช ืืืืืืช, ืืืื ืื ืื ืกื.** ืืื ืืฆืืข ืืฉืื ืืืจ: ืขืืจื ืกืืืจื, ืืื ืจืฉืช, ืืื ืฉืืฃ ืืื ืืืจ ืจืืื ืืช ืืงืื.
|
| 71 |
|
| 72 |
+
## ืืื ืืื ืืืืขื
|
| 73 |
|
| 74 |
+
- ืืฉืืื ืืกืงืืจื ืฉื ืงืื ืืกืืืื ืืคืืงืืช ืฉืืกืืจ ืื ืืฆืืช ืืืืฆื.
|
| 75 |
+
- ืืชืงื ื ืืชืื ืืืจื ืฉืืฉ ืื ืืืื ืืงืื-ืืืกืื ื ืืงืฉืื.
|
| 76 |
+
- ืขืืืื ืืืื ืขื ืืคืชื ืืฉืืืื ืืจื ื ืื ืืืื ืื ืื ืงืืื.
|
| 77 |
+
- ืืืืขื ืืชืื ืืื ืคืืชืื ืคื ืืื ืฉืืกืืจ ืื ืืงืจืื ื-API ืืืฆืื ื.
|
| 78 |
+
- ืืคืชืืื ืืืืจื ืขืืจืืช โ ืืืืื ืขืื ื ืืขืืจืืช ืืฉืคืื ืื ืืืื ืืขืืจืืช, ืืืงืื ืขืฆืื ื ืฉืืจ ืืื ืืืืช.
|
|
|
|
| 79 |
|
| 80 |
+
## ืื ืืื **ืื**, ืืื ืืกืืจ ืืขืฉืืช ืืืชื
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
| 82 |
+
- **ืืื ืื ืชืืืืฃ ืืืืื ืืืื** ืืฉืืืืช ืืจืืืืงืืืจื, ืืงืื ืฉืคืจืืก ืขื ืืจืื ืงืืฆืื, ืื ืืื ืืืจ ืฉืืืจืฉ ืืืืืืง ืืงืฉืจ ืืจืื ืืจืืฉ.
|
| 83 |
+
- **ืืกืืจ ืืฉืืื ืืช ืืงืื ืฉืื ืืืืฆืืจ ืืื ืฉืืื ืงืจื ืืืชื.** ืืื ืืืืฆืจ ืงืื ืฉื ืจืื ื ืืื ืืื ืจืฅ. ืื ืื ืชืงืื ื ืืืจื.
|
| 84 |
+
- **ืืื ืืืฆืื ืืืฉืงืื ืฉื ืกืคืจืืืช.** ืืชืืืืช ืฉื ืคืื ืงืฆืืืช ืฉืื ืงืืืืืช, ืคืจืืืจืื ืฉืื ืงืืืืื, ืืจืกืืืช ืฉืื ืงืืืืืช. ืชืืื ืืืืืง ืืื ืืชืืขืื.
|
| 85 |
+
- **ืืื ืืืืจ ืจืง Python ื-TypeScript.** ืืื ืฉืคื ืืืจืช ืืืืกืื ืืื ืืืื, ืืืชืืืืจ ืฉืืฆื ืื ืืืื ืืืืจื ื ืืื ืื ืืงืืื.
|
| 86 |
+
- **ืืฉ ืื ืชืืจืื ืืืข.** ืกืคืจืืืช ืืืืื ืฉืืฆืื ืืืจื ืฉืืืืื ื ืืกืฃ, ืืชืืืืช 2026, ืคืฉืื ืื ืงืืืืื ืืฉืืืื.
|
| 87 |
+
- **ืืื ืื ืืืื.** ืืื ืื ืืจืืฅ ืคืงืืืืช, ืื ืงืืจื ืงืืฆืื ืืื ืืืืง ืืช ืขืฆืื. ืืฉืืื ืืชื ืืืืช ืฉื ืกืืื ืฆืจืื ืืื ืืช ืืกืืืื ืชืฉืชืืช.
|
| 88 |
+
- **ืืื ืื ืฆืืื ืขื ืืืื ืืืืจ.** ืจืื ืืช ืคืจืง ืืืขืจืื โ ืืืืืงืืช ืฉืื ื ืขืฉื ืงืื ืืช ืืืื ืืืื ื ืืืื.
|
| 89 |
+
|
| 90 |
+
## ืืื ืืจืืฆืื
|
| 91 |
|
| 92 |
### Ollama
|
| 93 |
|
|
|
|
| 98 |
|
| 99 |
### llama.cpp
|
| 100 |
|
| 101 |
+
ืฉื ืืงืืืฅ ืืชืื ืืืืืจ ืืื `gemma-4-e4b-it.Q4_K_M.gguf`. ืืฉื ื ืฉืืจ ืืฉืื ืืื ืืื โ ืื ืืงืืืฅ ืฉื ืืืืื **ืืืืืื**, ืืื ืฉื ืืืื ืืืกืืก.
|
| 102 |
+
|
| 103 |
```bash
|
| 104 |
+
./llama-cli -m gemma-4-e4b-it.Q4_K_M.gguf \
|
| 105 |
-p "Write a Python function that parses ISO-8601 dates with timezones." \
|
| 106 |
--temp 0.2 --top-p 0.95 -n 1024
|
| 107 |
```
|
| 108 |
|
| 109 |
+
### ืคืืืชืื, ืืจื ืืืืจ ื-safetensors
|
| 110 |
|
| 111 |
```python
|
| 112 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
|
|
| 126 |
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 127 |
```
|
| 128 |
|
| 129 |
+
### ืคืจืืืจื ืืจืฆื ืืืืืฆืื
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
|
| 131 |
+
| ืคืจืืืจ | ืขืจื | ืืื |
|
| 132 |
+
|---|---|---|
|
| 133 |
+
| `temperature` | 0.2 | ืืฆืืจืชืืืช ื ืืืื. ืืงืื ืจืืฆืื ืชืฉืืื ืฆืคืืื, ืื ืืงืืจืืช |
|
| 134 |
+
| `top_p` | 0.95 | ืืขื ืืืื ืืืืืื ืืืฉืคืื, ืืื ืืืคืฉืจ ืืืืื ืืกืื ืื ืืชืืื |
|
| 135 |
+
| `max_new_tokens` | 1024 | ืืกืคืืง ืืจืื ืืคืื ืงืฆืืืช |
|
| 136 |
+
| `repetition_penalty` | 1.0 | ืงื ืก ืขื ืืืจืืช ืคืืืข ืืงืื โ ืืืื ืืฉืืืช ืืฉืชื ืื ืืืืจืื ืขื ืขืฆืื ืืืืื ื |
|
| 137 |
|
| 138 |
+
## ืคืจืืืคื ืืืขืจืืช ืืืืืืฅ โ ืื ืื ืฉืืฉื ื ืืื ืืจืื
|
| 139 |
|
| 140 |
+
ืืืื ืืืืื ืืื ืืืชื ืงืื **ืืจืื** ืืืชืจ ืืื ืืฉืืืจืืืื ืืืชื ืืืฉืื ืืืืืฉื ืฉืืืื ืืคื ื ืฉืืื ืืืชื ืฉืืจื. ืืื ืื ืืื ืงืืคืฅ ืืฉืจ ืืงืื โ ืงืื ืฉืืชืงืืคื ืื ืืคื ืขื ืืงืจื ืงืฆื, ืืื ืืืืงืืช ืืืื ืืืืจื.
|
| 141 |
|
| 142 |
+
ืืืฉืช ืืฉืืืื: ืืืืื ืืช ืืืขืื, ืืื ืืช ืืช ืืงืจื ืืงืฆื, ืืืชืื ืืช ืืงืื, ืืืชืื ืืืืงืืช, ืืืืืจ ืืื ืืช ืื ืืงืื ืื ืืืกื.
|
| 143 |
|
| 144 |
+
**ืืืืช ืืชืจืฉืืืช, ืื ืืืืื.** ืื ืจืฆื ืืฉืืืื ืืกืคืจืืช ืืื ืืจืฆื ืขื ืืคืจืืืคื ืืืจืฆื ืืืขืืื.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
|
| 146 |
+
### ืืคืจืืืคื ืขืฆืื โ ืืขืชืง ืืื ืฉืืื
|
| 147 |
|
| 148 |
```text
|
| 149 |
DEFINITIONS:
|
|
|
|
| 222 |
- regression check: No "production-ready" claims unless edge cases match limitations.
|
| 223 |
```
|
| 224 |
|
| 225 |
+
### ืืืืืช ืฉืืืืฉ ืขื ืืคืจืืืคื
|
| 226 |
|
| 227 |
```python
|
| 228 |
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
|
|
| 234 |
device_map="auto",
|
| 235 |
)
|
| 236 |
|
| 237 |
+
# ืืืืง ืืื ืืช ืืคืจืืืคื ืืืื ืืืืืืง ืฉืืืขืื
|
| 238 |
SYSTEM_PROMPT = """[paste the full prompt from the code block above]"""
|
| 239 |
|
| 240 |
messages = [
|
|
|
|
| 247 |
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
|
| 248 |
```
|
| 249 |
|
| 250 |
+
### ืืชืืืืช ืืคืฉืจืืืช
|
| 251 |
|
| 252 |
+
- ืจืืฆื ืจืง ืงืื ืืื ืืกืืจืื? ืืืืฃ ืืช `OUTPUT_FORMAT` ื-"Code blocks only".
|
| 253 |
+
- ืืื ื ืืื ืืกืงืืจืช ืงืื? ืืืกืฃ ื-`REQUIREMENTS` ืืจืืฉื ืฉืืคืื ืืืื ืืคืืจืื diff.
|
| 254 |
+
- ืฆืจืื ืจืง TypeScript? ืืืกืฃ ื-`REQUIREMENTS` ืฉืื ืชืฉืืื ืชืืื ื-TypeScript ืขื ืืืคืืกืื.
|
| 255 |
+
- ืขืืื ืขื ืงืื ืจืืืฉ ืืืืืชืืช? ืืืกืฃ ืกืขืืฃ ื-`OUTPUT_FORMAT` ืืฉื "Security Review".
|
| 256 |
|
| 257 |
+
## ืคืจืื ืืืืืื
|
| 258 |
|
| 259 |
+
| ืืืคืืื | ืขืจื |
|
| 260 |
+
|---|---|
|
| 261 |
+
| **ืืืื ืืกืืก** | [`unsloth/gemma-4-E4B-it`](https://huggingface.co/unsloth/gemma-4-E4B-it) |
|
| 262 |
+
| **ืฉืืื** | QLoRA โ ืืืื ืืืกืืก ื ืืขื ื-4 ืืื ืืืื ืืืืืื |
|
| 263 |
+
| **ืืกืืจืช** | Unsloth |
|
| 264 |
+
| **ืืืืจื** | NVIDIA RTX 5090 |
|
| 265 |
+
| **ืฉืืจืืช ืืืืื** | 38,314 |
|
| 266 |
+
| **ืฉืืจืืช ืืืื ืฉืืืืืงื ืืฆื** | 2,016 |
|
| 267 |
+
| **ืืืืงื** | 95% / 5%, seed 3407 |
|
| 268 |
+
| **ืืืคืจ-ืคืจืืืจืื, ืืื ืืขืืืช** | ืื ื ืืชืืื ืืื โ ืจืื ืืืขืจื ืืชืืช ืืืืื |
|
| 269 |
|
| 270 |
+
> **ืืื ืืกืจืื ืืื ืืกืคืจืื.** ืจืฉืืืืช ืืืืืื ืฉื ืืืืื ืืื ืฉืจืื ืืฉืชื ืืจืกืืืช ืฉืกืืชืจืืช ืื ืืช ืื ืืืืืง ืขื ืืจืืช ื-LoRA ืืขื ืฉืืจ ืืืืคืจ-ืคืจืืืจืื. ืื ืืคืฉืจ ืืืขืช ืืืืงืืจืืช ืฉืงืืืืื ืืืื ืืื ืืชืืจืช ืืช ืืืฉืงืืืืช ืฉืคืืจืกืื ืืื, ืืืื ืืฉืืจืืช ืืืื ืืืกืจื ืืืงืื ืืืืฉืืจ ืืืืืจืืืช ืืื ืขืืืื. ืื ืฉืื ื ืฉืืจ โ ืืืื ืืืกืืก, ืืืืืจื, ืืกืคืืจืช ืืฉืืจืืช โ ืืืคืืข ืืืืคื ืืื ืืฉื ื ืืืงืืจืืช, ืืกืคืืจืช ืืฉืืจืืช ืืฃ ื ืงืจืื ืืงืืืฅ ืกืืืืกืืืงื ืฉื ืืฆืจ ืขื ืืื ืืืืื ื ืขืฆืื.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 271 |
|
| 272 |
+
### ืืื ืืืจืื ืืืืื
|
| 273 |
|
| 274 |
+
| ื๏ฟฝ๏ฟฝืืจ | ืืืืช | ืชืืื |
|
| 275 |
+
|---|---|---|
|
| 276 |
+
| [`nvidia/OpenCodeInstruct`](https://huggingface.co/datasets/nvidia/OpenCodeInstruct) | 20,000 | Python โ ืจืง ืืืืืืืช ืฉืืงืื ืืื ืขืืจ ืืคืืืช 50% ืืืืืืงืืช ืฉืื |
|
| 277 |
+
| [`bleugreen/typescript-instruct`](https://huggingface.co/datasets/bleugreen/typescript-instruct) | 20,000 | TypeScript |
|
| 278 |
+
| ืกื ืืืืช ืฉื ืืชื ืืื | 330 | 165 ืืืืืช ืฉืืื-ืชืฉืืื, ืื ืืื ื ืืื ืคืขืืืื. ืขืืจืืช ืืื ืืืืช |
|
| 279 |
+
| **ืกื ืืื** | **40,330** | |
|
| 280 |
|
| 281 |
+
**ืืกืื ืื ืืื ืืขืืงืจ ืืื.** ืืงืืจ ื-Python ืืื ืงืืจืคืืก ืขื ืง. ืขืืจื ื ืขืืื ืืืชืื ื ืืคื ืืืื ืืื: ืืื ืืงืื ืืืืืื ืขืืจ ืืช ืืืืืงืืช ืฉื ืืชืื ืขืืืจื. ืืืืืืืช ืืื ืชืืฆืืืช ืืืืงื ื ืืจืงื, ืืืืืืืช ืฉืขืืจื ืคืืืช ืืืฆื ื ืืจืงื, ืืืคืืืืืืช ืืคื ืฉืืื ืืื ื ืืจืงื. ืื ืืืจื ืืืงืกื ื ืืชื ื-6,000 ืชืืืื.
|
| 282 |
|
| 283 |
+
ืื ืืืืชื ืืืืืื ืฉืืฉืคืืขื ืืื ืืจืื ืขื ืืชืืฆืื: ืืืืื ืขื ืืงืืจืคืืก ืืืื, ืืื ืกืื ืื, ืืืฆืจ ืืืื ืจืืขืฉ ืืืชืจ.
|
| 284 |
|
| 285 |
+
ืืื ืืคืืจืกื ืืืจืืืก ืืืืื ืฉื [`code-training-il`](https://huggingface.co/datasets/BrainboxAI/code-training-il).
|
| 286 |
|
| 287 |
+
## ืืขืจืื
|
| 288 |
|
| 289 |
+
**ืื ืจืฅ ืืืื ืืืืจ ืขื ืืืืื ืืื. ืืื ืฆืืื HumanEval, ืืื MBPP, ืืืื ืฉืื ืืกืคืจ ืฉืืคืฉืจ ืืืฉืืืช ืืืื ืืืืื ืืืจ.**
|
|
|
|
|
|
|
|
|
|
| 290 |
|
| 291 |
+
ืื ืฉืื ื ืขืฉื โ ืฉืชื ืืืืงืืช ืืื ืืืช, ืงืื ืืช:
|
| 292 |
|
| 293 |
+
| ืื ื ืืืง | ืืื ืืงืจืื | ืชืืฆืื |
|
| 294 |
+
|---|---|---|
|
| 295 |
+
| FizzBuzz, ืืจื ืืืืืช ืกืืื | 5 | 5 ืืชืื 5, ื-6 ืฆืขืืื, ืืื ืกืื ืชืืงืื |
|
| 296 |
+
| ืืืคืืฉ ืืื ืืจื ืขื 11 ืืงืจื ืงืฆื | 11 | 11 ืืชืื 11, ืืืื ืืืคืื ืืืคืืืืช ืืฉืืืืืช |
|
| 297 |
|
| 298 |
+
**ืืื ืืงืจืื ืืช ืื, ืืื ืืช.** ืืื 16 ืืงืจืื ืืกื ืืื, ืฉืฉื ื ืื ื ืืื ืืจืืฆื ืืื. ืืื ืงืืืฅ ืชืืฆืืืช, ืืื ืงืื ืืืืงื ืฉืคืืจืกื, ืืื ืืคืฉืจ ืืฉืืืจ ืืช ืื ืืืืืฅ. ืื ืืกืคืืง ืืื ืืืืจ "ืืืืื ืขืืื ืืื ืงืืจืก". ืื **ืื** ืืืื, ืืืกืืจ ืืืฉืืืช ืืืชื ืืืกืคืจืื ืฉื ืืืืืื ืืืจืื.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 299 |
|
| 300 |
+
ืืืื ืืืืชื ืืื ืขืืืื ืคืชืืื. ืื ืืืืฉืจ ืืื ืืจืืฅ, ืืชืืฆืื ืชืืคืืข ืืื.
|
| 301 |
|
| 302 |
+
## ืืืืืืช
|
|
|
|
| 303 |
|
| 304 |
+
- **ืืืื ืงืื.** ืืืืื ืืื ืืฉ ืืขืืืืช ืืฉืืืืช ืืจืืืืงืืืจื ืืืืงืฉืจ ืืจืื. ืื ืืืืืช, ืื ืืคืฉืจืืช.
|
| 305 |
+
- **ืฉืชื ืฉืคืืช.** ืืืง ื-Python ืื-TypeScript, ืืืฉ ืืื ืืฉืืจ.
|
| 306 |
+
- **ืืื ืฉืืืืฉ ืืืืื ืืืงืืคืกื.** ืืื ืืืืจ, ืืื ืื ืืจืืฅ. ืกืืื ืืืจืฉ ืขืืืืช ืืื ืืืจืฆืื.
|
| 307 |
+
- **ืชืืจืื ืืืข.** ืื ืฉืืฆื ืืืจื ืชืืืืช 2026 ืื ืงืืื ืืฉืืืื.
|
| 308 |
+
- **ืืื ืืืฆืื ืงืื ืฉื ืจืื ื ืืื.** ืชืืื ืืืจืืฅ ืืืืืืง.
|
| 309 |
+
- **ืืื ืืืื.** ืจืื ืืช ืคืจืง ืืืขืจืื.
|
| 310 |
+
- **ืืื ืืืืื ืืขื `unsloth/gemma-4-E4B-it`.** ืื ืืืืื ืฉื ืืืื ืืืกืืก ื ืืฆืืช ืื ืืื.
|
| 311 |
|
| 312 |
+
## ืืงืืฆืื ืืืืืืจืื
|
| 313 |
|
| 314 |
+
| ืืืืจ | ืื ืืฉ ืืคื ืื | ืืื ืื |
|
| 315 |
+
|---|---|---|
|
| 316 |
+
| [`BrainboxAI/code-il-E4B`](https://huggingface.co/BrainboxAI/code-il-E4B) | `gemma-4-e4b-it.Q4_K_M.gguf` (5.3 ื'ืืื) ืืืืจืืืก ืืื | Ollama, llama.cpp, LM Studio |
|
| 317 |
+
| [`BrainboxAI/code-il-E4B-safetensors`](https://huggingface.co/BrainboxAI/code-il-E4B-safetensors) | ืืฉืงืืืืช ืืืืืช ื-16 ืืื (16.0 ื'ืืื) | transformers, ืืืืฉื ืืืืื |
|
| 318 |
+
|
| 319 |
+
ืืืืืจ ืืืฉื ืื `gemma-4-e4b-it.BF16-mmproj.gguf` (0.99 ื'ืืื). ืืื ืจืืื ืืจืืืื ืฉื Gemma-4, ืฉื ืืืฅ ืจืง ืื ืจืืฆืื ืืืืื ืชืืื ืืช. ืืขืืืืช ืงืื ืืื ืื ืฆืืจื.
|
| 320 |
+
|
| 321 |
+
## ืจืืฉืืื
|
| 322 |
+
|
| 323 |
+
Apache 2.0. ืืืชืจ ืืืฉืชืืฉ, ืืฉื ืืช, ืืืคืืฅ ืืืืืืจ ื ืืืจืืช, ืขื ืืืืืก.
|
| 324 |
+
|
| 325 |
+
ืืื ืืืืื ืืขื [`unsloth/gemma-4-E4B-it`](https://huggingface.co/unsloth/gemma-4-E4B-it), ืืืื ืืชื ืืื ืฉื ืืืื ืืืกืืก ืืืื ืื ืขื ืืืืื ืืื. ืืืื ืืืกืืก ืืคืืจืกื ืชืืช Apache 2.0 ืืืคื ื ืื ืื [ืชื ืื ืืฉืืืืฉ ืฉื Gemma](https://ai.google.dev/gemma/docs/gemma_4_license). ืืืื ืืงืจืื ืืืชื ืืคื ื ืฉืืืืฉ ืืกืืจื.
|
| 326 |
+
|
| 327 |
+
ืืืืืจ ืืืืืื ืืฉ ืจืืฉืืื ืืช ืืฉืื, ืฉื ืืืงืืจืืช ืฉืืื ื ืืื ื ืื ื. ืจืื ืืช ืืจืืืก ืืืืื.
|
| 328 |
+
|
| 329 |
+
## ืฆืืืื
|
| 330 |
|
| 331 |
```bibtex
|
| 332 |
@misc{elyasi2026codeil,
|
| 333 |
+
title = {Code-IL E4B (bx-code-nogah): A Small, On-Device Coding Assistant for Private Environments},
|
| 334 |
author = {Elyasi, Netanel},
|
| 335 |
year = {2026},
|
| 336 |
publisher = {BrainboxAI},
|
|
|
|
| 339 |
}
|
| 340 |
```
|
| 341 |
|
| 342 |
+
## ืื ืื ื ืืช ืื
|
| 343 |
|
| 344 |
+
ื ืื ื ืขื ืืื [**ื ืชื ืื ืืืืืกื**](https://huggingface.co/BrainboxAI), ืืืืกื [BrainboxAI](https://brainboxai.io) โ ืกืืืืื ืืฉืจืืื ืืืื ื ืืืืืืชืืช ืืืฉืืืืช, ืฉืืื ื ืืืืืื ืงืื ืื, ืคืจืืืื ืืืชืืืื.
|
| 345 |
|
| 346 |
+
ืืืืื ืื ืืืื ืงืื ืขื ืืกืืก ืืงืื ืืคืจืื ืฉื ืืืืจื ืฉืื: [netanele@brainboxai.io](mailto:netanele@brainboxai.io).
|
|
|
|
|
|
|
| 347 |
|
| 348 |
+
*ืืืง ืืืฉืคืืช ืืืืืืื ืฉื BrainboxAI ืฉืจืฆืื ืขื ืืืืืจื ืฉืื. ืจืื ืื [`law-il-E2B`](https://huggingface.co/BrainboxAI/law-il-E2B) (ืืฉืคื) ื-[`cyber-analyst-4B`](https://huggingface.co/BrainboxAI/cyber-analyst-4B) (ืกืืืืจ).*
|