Text Generation
Transformers
Safetensors
grt
code
python
tool-calling
function-calling
rotational-transformer
givens-rotations
clifford-algebra
matterwave
Instructions to use MatterWaveAI/grt-30m-code with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use MatterWaveAI/grt-30m-code with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="MatterWaveAI/grt-30m-code")# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("MatterWaveAI/grt-30m-code", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use MatterWaveAI/grt-30m-code with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "MatterWaveAI/grt-30m-code" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "MatterWaveAI/grt-30m-code", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/MatterWaveAI/grt-30m-code
- SGLang
How to use MatterWaveAI/grt-30m-code 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 "MatterWaveAI/grt-30m-code" \ --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": "MatterWaveAI/grt-30m-code", "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 "MatterWaveAI/grt-30m-code" \ --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": "MatterWaveAI/grt-30m-code", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use MatterWaveAI/grt-30m-code with Docker Model Runner:
docker model run hf.co/MatterWaveAI/grt-30m-code
Release MatterWaveAI/grt-30m-code (zero nn.Linear, 8-way classical geometric optimizer ablation)
6f6a759 verified Download optimizer_ablation_results.json from MatterWaveAI/grt-30m-code: direct link, hf CLI and curl.
- Browser
- Download file 4.91 kB
-
https://huggingface.co/MatterWaveAI/grt-30m-code/resolve/main/optimizer_ablation_results.json
- Command line
-
hf download hf://MatterWaveAI/grt-30m-code/optimizer_ablation_results.json
-
curl -L -o optimizer_ablation_results.json https://huggingface.co/MatterWaveAI/grt-30m-code/resolve/main/optimizer_ablation_results.json
4.91 kB
| { | |
| "model_id": "MatterWaveAI/grt-30m-code", | |
| "architecture": "GRTForCausalLM", | |
| "pipeline_tag": "text-generation", | |
| "base_model": "MatterWaveAI/grt-30m", | |
| "parameters": 29305498, | |
| "trainable_parameters": 811907, | |
| "trainable_scope": "blocks[11:12] + final_norm + <|im_end|> phase calibration", | |
| "nn_linear_layers": 0, | |
| "nn_conv1d_layers": 0, | |
| "steps": 200, | |
| "batch_size": 16, | |
| "learning_rate": 0.0035, | |
| "seed": 42, | |
| "best_optimizer": "huygens_adam", | |
| "before_training": { | |
| "val_loss": 4.0061, | |
| "val_ppl": 54.9316, | |
| "corrupted_prompt_val_loss": 4.9059, | |
| "code_conditioning_margin": 0.8998, | |
| "code_token_acc": 46.8, | |
| "tool_call_token_acc": 36.44, | |
| "json_schema_match_rate": 0.0, | |
| "python_ast_syntax_validity": 0.0 | |
| }, | |
| "optimizers": { | |
| "pair_adam": { | |
| "optimizer_type": "pair_adam", | |
| "variant": "PairAdam (Factored Toroidal Baseline)", | |
| "final_train_loss": 1.0994, | |
| "val_loss": 0.6018, | |
| "val_ppl": 1.8253, | |
| "code_token_acc": 97.6, | |
| "tool_call_token_acc": 90.69, | |
| "json_schema_match_rate": 83.33, | |
| "python_ast_syntax_validity": 50.0, | |
| "code_conditioning_margin": 1.6597, | |
| "corrupted_prompt_val_loss": 2.2614, | |
| "wall_time_s": 278.8 | |
| }, | |
| "hessian_pair_adam": { | |
| "optimizer_type": "hessian_pair_adam", | |
| "variant": "HessianPairAdam (Sophia-Torus Cauchy-Riemann Curvature)", | |
| "final_train_loss": 1.2171, | |
| "val_loss": 0.7292, | |
| "val_ppl": 2.0734, | |
| "code_token_acc": 97.87, | |
| "tool_call_token_acc": 87.33, | |
| "json_schema_match_rate": 66.67, | |
| "python_ast_syntax_validity": 58.33, | |
| "code_conditioning_margin": 1.4236, | |
| "corrupted_prompt_val_loss": 2.1528, | |
| "wall_time_s": 117.82 | |
| }, | |
| "polar_natural_adam": { | |
| "optimizer_type": "polar_natural_adam", | |
| "variant": "PolarNaturalAdam (Riemannian Polar Metric + Cauchy-Riemann Hessian)", | |
| "final_train_loss": 1.2463, | |
| "val_loss": 0.7891, | |
| "val_ppl": 2.2015, | |
| "code_token_acc": 96.93, | |
| "tool_call_token_acc": 86.34, | |
| "json_schema_match_rate": 66.67, | |
| "python_ast_syntax_validity": 50.0, | |
| "code_conditioning_margin": 1.4294, | |
| "corrupted_prompt_val_loss": 2.2185, | |
| "wall_time_s": 108.18 | |
| }, | |
| "cyclide_adam": { | |
| "optimizer_type": "cyclide_adam", | |
| "variant": "CyclideAdam (Dupin Cyclide Conic Metric + Deviatoric Twist)", | |
| "final_train_loss": 1.1564, | |
| "val_loss": 0.64, | |
| "val_ppl": 1.8965, | |
| "code_token_acc": 98.93, | |
| "tool_call_token_acc": 89.5, | |
| "json_schema_match_rate": 75.0, | |
| "python_ast_syntax_validity": 83.33, | |
| "code_conditioning_margin": 1.5727, | |
| "corrupted_prompt_val_loss": 2.2127, | |
| "wall_time_s": 141.92 | |
| }, | |
| "symplectic_adam": { | |
| "optimizer_type": "symplectic_adam", | |
| "variant": "SymplecticAdam (Gromov-Liouville Non-Squeezing Floor + Casimir Cavity)", | |
| "final_train_loss": 1.2764, | |
| "val_loss": 0.8445, | |
| "val_ppl": 2.3268, | |
| "code_token_acc": 97.6, | |
| "tool_call_token_acc": 81.39, | |
| "json_schema_match_rate": 50.0, | |
| "python_ast_syntax_validity": 50.0, | |
| "code_conditioning_margin": 1.5766, | |
| "corrupted_prompt_val_loss": 2.421, | |
| "wall_time_s": 138.1 | |
| }, | |
| "gyric_adam": { | |
| "optimizer_type": "gyric_adam", | |
| "variant": "GyricAdam (Poinsot-Lande Gyroscopic Projection + Thomas Precession)", | |
| "final_train_loss": 1.1168, | |
| "val_loss": 0.6216, | |
| "val_ppl": 1.8619, | |
| "code_token_acc": 98.53, | |
| "tool_call_token_acc": 87.92, | |
| "json_schema_match_rate": 75.0, | |
| "python_ast_syntax_validity": 50.0, | |
| "code_conditioning_margin": 1.5745, | |
| "corrupted_prompt_val_loss": 2.1961, | |
| "wall_time_s": 138.83 | |
| }, | |
| "huygens_adam": { | |
| "optimizer_type": "huygens_adam", | |
| "variant": "HuygensAdam (Circular Order Parameter R_p Splay Torque)", | |
| "final_train_loss": 1.0882, | |
| "val_loss": 0.5672, | |
| "val_ppl": 1.7632, | |
| "code_token_acc": 99.07, | |
| "tool_call_token_acc": 90.69, | |
| "json_schema_match_rate": 83.33, | |
| "python_ast_syntax_validity": 91.67, | |
| "code_conditioning_margin": 1.5152, | |
| "corrupted_prompt_val_loss": 2.0824, | |
| "wall_time_s": 138.63 | |
| }, | |
| "huygens_symplectic_adam": { | |
| "optimizer_type": "huygens_symplectic_adam", | |
| "variant": "HuygensSymplecticAdam (Two-Stage Huygens -> Symplectic Curriculum)", | |
| "final_train_loss": 1.1819, | |
| "val_loss": 0.6537, | |
| "val_ppl": 1.9227, | |
| "code_token_acc": 98.27, | |
| "tool_call_token_acc": 89.31, | |
| "json_schema_match_rate": 75.0, | |
| "python_ast_syntax_validity": 66.67, | |
| "code_conditioning_margin": 1.4057, | |
| "corrupted_prompt_val_loss": 2.0594, | |
| "wall_time_s": 122.82 | |
| } | |
| } | |
| } | |