Text Generation
Transformers
Safetensors
English
gabgpt
gift-of-gab
small-language-model
trained-from-scratch
byte-level
looped-transformer
browser
conversational
custom_code
Instructions to use gszauer/Gab10M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use gszauer/Gab10M with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="gszauer/Gab10M", trust_remote_code=True) messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("gszauer/Gab10M", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use gszauer/Gab10M with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "gszauer/Gab10M" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "gszauer/Gab10M", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/gszauer/Gab10M
- SGLang
How to use gszauer/Gab10M 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 "gszauer/Gab10M" \ --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": "gszauer/Gab10M", "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 "gszauer/Gab10M" \ --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": "gszauer/Gab10M", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use gszauer/Gab10M with Docker Model Runner:
docker model run hf.co/gszauer/Gab10M
File size: 2,302 Bytes
04036d0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 | {
"name": "Gab10M",
"architecture": "gab10m",
"vocabSize": 256,
"featureDim": 256,
"numHeads": 4,
"ropeBase": 10000,
"numPreludeBlocks": 2,
"numCoreBlocks": 9,
"numCodaBlocks": 2,
"loopCount": 2,
"mlpHiddenDim": 680,
"convKernel": 4,
"contextLength": 2048,
"derived": {
"headDim": 64,
"parameterCount": 10302720,
"uniqueBlocks": 13,
"effectiveDepth": 22
},
"looping": {
"schedule": "prelude once, core loopCount times with shared weights, coda once",
"injection": "additive: the prelude output is added to the residual stream at the start of every core pass after the first",
"kvCache": "one slot per layer pass (effectiveDepth slots), so each core block has loopCount slots; each slot also holds that pass's two canon histories"
},
"chat": {
"user": "<|u|>",
"assistant": "<|a|>",
"end": "<|e|>",
"endOfText": "<|endoftext|>",
"encoding": "markers are plain UTF-8 text spelled out byte by byte; no reserved ids",
"template": "<|u|>{user}<|e|><|a|>{assistant}<|e|>, repeated for each turn pair",
"structure": "one or more user/assistant turn pairs, starting with the user; no system prompt",
"generationPrompt": "the conversation so far followed by <|a|>; the reply ends at <|e|>"
},
"weights": {
"format": "self-describing GAB_ file: 12-byte preamble (magic GAB_, version, header length), JSON header, then little-endian float32 weights in this order",
"matrixLayout": "row-major [in, out]; activations are multiplied as x @ W",
"order": [
"tokenEmbeddings [vocabSize, featureDim] (tied: also the output projection, transposed)",
"for each unique block, prelude then core then coda: attentionNorm.gamma [1, featureDim], canonA.weight [convKernel, featureDim] (row j weighs the input j positions back)",
" for each head: learnedQ [featureDim, headDim], learnedK [featureDim, headDim], learnedV [featureDim, headDim], learnedO [headDim, featureDim], queryNorm.gamma [1, headDim], keyNorm.gamma [1, headDim]",
" mlpNorm.gamma [1, featureDim], canonC.weight [convKernel, featureDim], mlp.learnedGate [featureDim, mlpHiddenDim], mlp.learnedUp [featureDim, mlpHiddenDim], mlp.learnedDown [mlpHiddenDim, featureDim]",
"finalNorm.gamma [1, featureDim]"
]
}
}
|