Text Generation
Transformers
Safetensors
English
cma
custom_code
causal-lm
small-language-model
generalist
4k-tokenizer
Instructions to use User01110/CMA-8M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use User01110/CMA-8M with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="User01110/CMA-8M", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("User01110/CMA-8M", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use User01110/CMA-8M with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "User01110/CMA-8M" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "User01110/CMA-8M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/User01110/CMA-8M
- SGLang
How to use User01110/CMA-8M 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 "User01110/CMA-8M" \ --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": "User01110/CMA-8M", "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 "User01110/CMA-8M" \ --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": "User01110/CMA-8M", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use User01110/CMA-8M with Docker Model Runner:
docker model run hf.co/User01110/CMA-8M
Remove leaderboard framing from model card
Browse files- README.md +9 -22
- assets/training-trajectory.svg +29 -27
README.md
CHANGED
|
@@ -25,7 +25,7 @@ A compact generalist language model built to study channel mixing at small scale
|
|
| 25 |
|
| 26 |
**7.85M parameters | 1,024-token context | 4,096-token vocabulary | Safetensors**
|
| 27 |
|
| 28 |
-
**
|
| 29 |
|
| 30 |
</div>
|
| 31 |
|
|
@@ -33,9 +33,9 @@ A compact generalist language model built to study channel mixing at small scale
|
|
| 33 |
|
| 34 |
## Overview
|
| 35 |
|
| 36 |
-
CMA 8M is
|
| 37 |
-
|
| 38 |
-
|
| 39 |
|
| 40 |
| Detail | Value |
|
| 41 |
|---|---|
|
|
@@ -54,7 +54,7 @@ CMA in the feed-forward position. For every token, CMA reshapes the 288-channel
|
|
| 54 |
hidden state into **12 chunks of 24**, routes between those chunks with **3
|
| 55 |
heads**, applies a signed blend and SiLU gate, then projects back to 288 channels.
|
| 56 |
|
| 57 |
-
 + PIQA + ArithMark-2) / 4
|
| 164 |
-
```
|
| 165 |
-
|
| 166 |
-
This matches the Open SLM Leaderboard formula. The checkpoint was selected at
|
| 167 |
-
step 40,000 for its **36.21%** average.
|
| 168 |
|
| 169 |
### WikiText-103 validation
|
| 170 |
|
|
@@ -176,5 +165,3 @@ step 40,000 for its **36.21%** average.
|
|
| 176 |
| Scored tokens | 358,911 |
|
| 177 |
| Normalized UTF-8 bytes | 1,145,226 |
|
| 178 |
|
| 179 |
-
WikiText normalized BPB is reported separately and was **not** used for
|
| 180 |
-
checkpoint selection.
|
|
|
|
| 25 |
|
| 26 |
**7.85M parameters | 1,024-token context | 4,096-token vocabulary | Safetensors**
|
| 27 |
|
| 28 |
+
**40,000 training steps | 20.97B trained tokens**
|
| 29 |
|
| 30 |
</div>
|
| 31 |
|
|
|
|
| 33 |
|
| 34 |
## Overview
|
| 35 |
|
| 36 |
+
CMA 8M is a **7,849,161-parameter** Channel-Mixing Attention generalist
|
| 37 |
+
causal language model trained for 40,000 steps. It pairs a compact nine-layer
|
| 38 |
+
architecture with the unmodified native GPT-S 4,096-token tokenizer.
|
| 39 |
|
| 40 |
| Detail | Value |
|
| 41 |
|---|---|
|
|
|
|
| 54 |
hidden state into **12 chunks of 24**, routes between those chunks with **3
|
| 55 |
heads**, applies a signed blend and SiLU gate, then projects back to 288 channels.
|
| 56 |
|
| 57 |
+

|
| 58 |
|
| 59 |
## Quick start
|
| 60 |
|
|
|
|
| 133 |
|
| 134 |
## Training trajectory
|
| 135 |
|
| 136 |
+
Across the logged validation checkpoints, normalized bits per byte (BPB)
|
| 137 |
+
decreased from **3.7838** to **1.3520**, while validation loss decreased from
|
| 138 |
+
**8.3686** to **2.9903**.
|
| 139 |
|
| 140 |
+

|
| 141 |
|
| 142 |
## Evaluation
|
| 143 |
|
|
|
|
| 154 |
| ARC-Challenge | **23.29%** |
|
| 155 |
| PIQA | **58.22%** |
|
| 156 |
| ArithMark-2 | **29.12%** |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157 |
|
| 158 |
### WikiText-103 validation
|
| 159 |
|
|
|
|
| 165 |
| Scored tokens | 358,911 |
|
| 166 |
| Normalized UTF-8 bytes | 1,145,226 |
|
| 167 |
|
|
|
|
|
|
assets/training-trajectory.svg
CHANGED
|
|
|
|