Text Generation
GGUF
English
flatbuild
language-model
conversational
flatseek
flatrun
trained-from-scratch
rope
rmsnorm
swiglu
gqa
q4_0
Instructions to use flatseek/flatbot-mini-35M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use flatseek/flatbot-mini-35M 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 flatseek/flatbot-mini-35M # Run inference directly in the terminal: llama cli -hf flatseek/flatbot-mini-35M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf flatseek/flatbot-mini-35M # Run inference directly in the terminal: llama cli -hf flatseek/flatbot-mini-35M
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf flatseek/flatbot-mini-35M # Run inference directly in the terminal: ./build/bin/llama cli -hf flatseek/flatbot-mini-35M
Use Docker
docker model run hf.co/flatseek/flatbot-mini-35M
- Llama new
- LM Studio
- Jan
- vLLM
How to use flatseek/flatbot-mini-35M with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "flatseek/flatbot-mini-35M" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "flatseek/flatbot-mini-35M", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/flatseek/flatbot-mini-35M
- Ollama
How to use flatseek/flatbot-mini-35M with Ollama:
ollama run hf.co/flatseek/flatbot-mini-35M
- Unsloth Desktop
- Docker Model Runner
How to use flatseek/flatbot-mini-35M with Docker Model Runner:
docker model run hf.co/flatseek/flatbot-mini-35M
- Lemonade
How to use flatseek/flatbot-mini-35M with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull flatseek/flatbot-mini-35M
Run and chat with the model
lemonade run user.flatbot-mini-35M-{{QUANT_TAG}}List all available models
lemonade list
- Atomic Chat
Update README.md
Browse files
README.md
CHANGED
|
@@ -19,17 +19,17 @@ tags:
|
|
| 19 |
pipeline_tag: text-generation
|
| 20 |
---
|
| 21 |
|
| 22 |
-
# Flatbot-Mini-
|
| 23 |
|
| 24 |
**Try it online:** https://chat.flatseek.io
|
| 25 |
|
| 26 |
-
Flatbot-
|
| 27 |
|
| 28 |
-
It is
|
| 29 |
|
| 30 |
-
The original FP32 checkpoint is approximately **135 MB**, while the **Q4_0 GGUF**
|
| 31 |
|
| 32 |
-
> **Experimental model:** Flatbot-
|
| 33 |
|
| 34 |
---
|
| 35 |
|
|
@@ -44,7 +44,7 @@ The original FP32 checkpoint is approximately **135 MB**, while the **Q4_0 GGUF*
|
|
| 44 |
| Attention | Grouped Query Attention (16 Query Heads / 4 KV Heads) |
|
| 45 |
| Weight Tying | Yes |
|
| 46 |
| Context Length | 512 |
|
| 47 |
-
| Parameters |
|
| 48 |
|
| 49 |
---
|
| 50 |
|
|
@@ -66,7 +66,7 @@ rope_theta = 10000
|
|
| 66 |
|
| 67 |
# Training
|
| 68 |
|
| 69 |
-
|
| 70 |
|
| 71 |
## Dataset
|
| 72 |
|
|
@@ -77,7 +77,7 @@ The model was trained entirely from random initialization using **FlatBuild**.
|
|
| 77 |
| Validation Split | 5% |
|
| 78 |
| Context Length | 512 tokens |
|
| 79 |
|
| 80 |
-
The
|
| 81 |
|
| 82 |
---
|
| 83 |
|
|
@@ -98,25 +98,28 @@ The training dataset consists of approximately **10,000** conversational example
|
|
| 98 |
|
| 99 |
# Quantization
|
| 100 |
|
| 101 |
-
This release is
|
| 102 |
|
| 103 |
| Item | Value |
|
| 104 |
|---|---:|
|
| 105 |
-
| Original Format | FP32 |
|
| 106 |
| Original Size | ~135 MB |
|
| 107 |
| Quantization | GGUF Q4_0 |
|
| 108 |
-
|
|
|
|
|
|
|
|
| 109 |
|
| 110 |
---
|
| 111 |
|
| 112 |
# Features
|
| 113 |
|
| 114 |
- Trained entirely from scratch
|
| 115 |
-
-
|
|
|
|
| 116 |
- Native chat template
|
| 117 |
- RoPE positional embeddings
|
| 118 |
-
- RMSNorm
|
| 119 |
-
- SwiGLU
|
| 120 |
- Grouped Query Attention (GQA)
|
| 121 |
- Weight-tied embeddings
|
| 122 |
- SafeTensors export
|
|
@@ -133,12 +136,12 @@ This release is provided as a **GGUF Q4_0** model optimized for lightweight infe
|
|
| 133 |
```bash
|
| 134 |
pip install flatbuild
|
| 135 |
|
| 136 |
-
flatbuild train configs/flatbot-mini-
|
| 137 |
|
| 138 |
flatbuild export \
|
| 139 |
-
outputs/flatbot-mini-
|
| 140 |
--format gguf \
|
| 141 |
-
--output flatbot-mini-
|
| 142 |
```
|
| 143 |
|
| 144 |
---
|
|
@@ -146,7 +149,7 @@ flatbuild export \
|
|
| 146 |
## LM Studio
|
| 147 |
|
| 148 |
```bash
|
| 149 |
-
lms import flatbot-mini-
|
| 150 |
```
|
| 151 |
|
| 152 |
---
|
|
@@ -157,7 +160,7 @@ lms import flatbot-mini-11M-Q4_0.gguf
|
|
| 157 |
pip install flatrun
|
| 158 |
|
| 159 |
flatrun chat \
|
| 160 |
-
--model flatbot-mini-
|
| 161 |
--temp 0.2
|
| 162 |
```
|
| 163 |
|
|
@@ -171,30 +174,30 @@ Loaded model in 0.03 s
|
|
| 171 |
You: Who are you?
|
| 172 |
|
| 173 |
Assistant:
|
| 174 |
-
I'm Flatbot, a conversational AI assistant trained from scratch using the Flatseek ecosystem. I'm
|
| 175 |
```
|
| 176 |
|
| 177 |
---
|
| 178 |
|
| 179 |
# Purpose
|
| 180 |
|
| 181 |
-
Flatbot-
|
| 182 |
|
| 183 |
-
1. Build conversational
|
| 184 |
2. Train a tokenizer
|
| 185 |
-
3. Configure a
|
| 186 |
4. Train entirely from scratch
|
| 187 |
5. Export SafeTensors checkpoints
|
| 188 |
6. Convert to GGUF
|
| 189 |
7. Quantize to Q4_0
|
| 190 |
8. Run efficient local inference with FlatRun
|
| 191 |
|
| 192 |
-
The entire workflow is fully reproducible on consumer hardware and showcases how modern language models can be
|
| 193 |
|
| 194 |
---
|
| 195 |
|
| 196 |
# Limitations
|
| 197 |
|
| 198 |
-
Flatbot-
|
| 199 |
|
| 200 |
-
Its primary
|
|
|
|
| 19 |
pipeline_tag: text-generation
|
| 20 |
---
|
| 21 |
|
| 22 |
+
# Flatbot-Mini-35M-Q4_0
|
| 23 |
|
| 24 |
**Try it online:** https://chat.flatseek.io
|
| 25 |
|
| 26 |
+
Flatbot-Mini-35M-Q4_0 is the flagship compact conversational model of the **Flatseek** ecosystem.
|
| 27 |
|
| 28 |
+
It is a **34.9 million parameter** decoder-only Transformer trained entirely from scratch using **FlatBuild**, exported to **GGUF**, quantized to **Q4_0**, and optimized for fast CPU inference with **FlatRun**.
|
| 29 |
|
| 30 |
+
The original FP32 checkpoint is approximately **135 MB**, while the quantized **Q4_0 GGUF** model significantly reduces memory and storage requirements, making it well suited for efficient local deployment on consumer hardware.
|
| 31 |
|
| 32 |
+
> **Experimental model:** Flatbot-Mini-35M is designed for research, education, and experimentation. Although considerably more capable than Flatbot-Micro-4M, it may still generate hallucinations, factual inaccuracies, or inconsistent responses.
|
| 33 |
|
| 34 |
---
|
| 35 |
|
|
|
|
| 44 |
| Attention | Grouped Query Attention (16 Query Heads / 4 KV Heads) |
|
| 45 |
| Weight Tying | Yes |
|
| 46 |
| Context Length | 512 |
|
| 47 |
+
| Parameters | 34.9M |
|
| 48 |
|
| 49 |
---
|
| 50 |
|
|
|
|
| 66 |
|
| 67 |
# Training
|
| 68 |
|
| 69 |
+
Flatbot-Mini-35M was trained entirely from random initialization using **FlatBuild** without relying on pretrained weights.
|
| 70 |
|
| 71 |
## Dataset
|
| 72 |
|
|
|
|
| 77 |
| Validation Split | 5% |
|
| 78 |
| Context Length | 512 tokens |
|
| 79 |
|
| 80 |
+
The dataset contains approximately **10,000** multi-turn conversational examples covering greetings, question answering, explanations, recommendations, coding assistance, reasoning, and general-purpose assistant interactions.
|
| 81 |
|
| 82 |
---
|
| 83 |
|
|
|
|
| 98 |
|
| 99 |
# Quantization
|
| 100 |
|
| 101 |
+
This release is distributed as a **GGUF Q4_0** model for efficient inference.
|
| 102 |
|
| 103 |
| Item | Value |
|
| 104 |
|---|---:|
|
| 105 |
+
| Original Format | FP32 SafeTensors |
|
| 106 |
| Original Size | ~135 MB |
|
| 107 |
| Quantization | GGUF Q4_0 |
|
| 108 |
+
| Parameters | 34.9M |
|
| 109 |
+
| Optimized For | CPU inference |
|
| 110 |
+
| Compatible Runtimes | FlatRun, llama.cpp, LM Studio, Ollama (GGUF) |
|
| 111 |
|
| 112 |
---
|
| 113 |
|
| 114 |
# Features
|
| 115 |
|
| 116 |
- Trained entirely from scratch
|
| 117 |
+
- 34.9M parameter Transformer
|
| 118 |
+
- Custom tokenizer (1,024 vocabulary)
|
| 119 |
- Native chat template
|
| 120 |
- RoPE positional embeddings
|
| 121 |
+
- RMSNorm normalization
|
| 122 |
+
- SwiGLU feed-forward network
|
| 123 |
- Grouped Query Attention (GQA)
|
| 124 |
- Weight-tied embeddings
|
| 125 |
- SafeTensors export
|
|
|
|
| 136 |
```bash
|
| 137 |
pip install flatbuild
|
| 138 |
|
| 139 |
+
flatbuild train configs/flatbot-mini-35M.yaml
|
| 140 |
|
| 141 |
flatbuild export \
|
| 142 |
+
outputs/flatbot-mini-35M/*/checkpoint/final \
|
| 143 |
--format gguf \
|
| 144 |
+
--output flatbot-mini-35M
|
| 145 |
```
|
| 146 |
|
| 147 |
---
|
|
|
|
| 149 |
## LM Studio
|
| 150 |
|
| 151 |
```bash
|
| 152 |
+
lms import flatbot-mini-35M-Q4_0.gguf
|
| 153 |
```
|
| 154 |
|
| 155 |
---
|
|
|
|
| 160 |
pip install flatrun
|
| 161 |
|
| 162 |
flatrun chat \
|
| 163 |
+
--model flatbot-mini-35M-Q4_0.gguf \
|
| 164 |
--temp 0.2
|
| 165 |
```
|
| 166 |
|
|
|
|
| 174 |
You: Who are you?
|
| 175 |
|
| 176 |
Assistant:
|
| 177 |
+
I'm Flatbot, a conversational AI assistant trained from scratch using the Flatseek ecosystem. I'm here to answer questions, explain concepts, and help with everyday tasks.
|
| 178 |
```
|
| 179 |
|
| 180 |
---
|
| 181 |
|
| 182 |
# Purpose
|
| 183 |
|
| 184 |
+
Flatbot-Mini-35M demonstrates the complete **Flatseek AI development pipeline**:
|
| 185 |
|
| 186 |
+
1. Build a conversational dataset
|
| 187 |
2. Train a tokenizer
|
| 188 |
+
3. Configure a Transformer architecture
|
| 189 |
4. Train entirely from scratch
|
| 190 |
5. Export SafeTensors checkpoints
|
| 191 |
6. Convert to GGUF
|
| 192 |
7. Quantize to Q4_0
|
| 193 |
8. Run efficient local inference with FlatRun
|
| 194 |
|
| 195 |
+
The entire workflow is fully reproducible on consumer hardware and showcases how modern language models can be trained, exported, quantized, and deployed without relying on proprietary foundation models.
|
| 196 |
|
| 197 |
---
|
| 198 |
|
| 199 |
# Limitations
|
| 200 |
|
| 201 |
+
Flatbot-Mini-35M is still a compact language model compared with modern foundation models containing billions of parameters. While it offers substantially stronger conversational ability than the earlier Flatbot-Micro-4M demonstration model, it remains limited in factual knowledge, complex reasoning, multilingual capability, and long-context understanding.
|
| 202 |
|
| 203 |
+
Its primary purpose is to demonstrate an end-to-end open-source training, quantization, and inference pipeline built entirely within the Flatseek ecosystem.
|