Any-to-Any
MLX
Safetensors
gemma4
mlx-vlm
rlcd
multimodal
classification
parallel-inference
image-text-to-text
audio
video
4-bit precision
Instructions to use larkooo/gemma-e2b-rlcd with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use larkooo/gemma-e2b-rlcd with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir gemma-e2b-rlcd larkooo/gemma-e2b-rlcd
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
| { | |
| "model": "models/gemma-4-e2b-it-4bit", | |
| "status": "cache_numerics_and_timing_probe_not_quality_benchmark", | |
| "results": [ | |
| { | |
| "state": "text", | |
| "prefix_tokens": 50, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 1.15625, | |
| "max_probability_delta": 4.3084425290461535e-05, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 1.65625, | |
| "max_probability_delta": 5.5105002298472456e-05, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 0.30837187499855645, | |
| 0.29970408300869167, | |
| 0.3233241660054773 | |
| ], | |
| "cached_batched": [ | |
| 0.38295441700029187, | |
| 0.3663168329803739, | |
| 0.4002443749923259 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 0.30837187499855645, | |
| "cached_batched": 0.38295441700029187 | |
| }, | |
| "speedup": 0.8052443353808384 | |
| }, | |
| { | |
| "state": "image", | |
| "prefix_tokens": 295, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 1.8125, | |
| "max_probability_delta": 0.08184582251343775, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 1.5, | |
| "max_probability_delta": 0.07954127470411075, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 2.9452710000041407, | |
| 2.088067457982106, | |
| 2.104747708013747 | |
| ], | |
| "cached_batched": [ | |
| 0.8102280829916708, | |
| 0.8053822079964448, | |
| 0.8086074579914566 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 2.104747708013747, | |
| "cached_batched": 0.8086074579914566 | |
| }, | |
| "speedup": 2.6029288837402547 | |
| }, | |
| { | |
| "state": "speech", | |
| "prefix_tokens": 86, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 0.6875, | |
| "max_probability_delta": 0.1209109506311471, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 0.6875, | |
| "max_probability_delta": 0.13835689201493084, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 0.390588584006764, | |
| 0.3957693339907564, | |
| 0.3943458329886198 | |
| ], | |
| "cached_batched": [ | |
| 0.3282808749936521, | |
| 0.3258304579940159, | |
| 0.3353305420023389 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 0.3943458329886198, | |
| "cached_batched": 0.3282808749936521 | |
| }, | |
| "speedup": 1.2012452233053332 | |
| }, | |
| { | |
| "state": "video", | |
| "prefix_tokens": 328, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 0.40625, | |
| "max_probability_delta": 0.016064510780345466, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 0.5, | |
| "max_probability_delta": 0.022443869382056524, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 1.817464583000401, | |
| 1.9493792499997653, | |
| 1.6563971669820603 | |
| ], | |
| "cached_batched": [ | |
| 0.7273326250142418, | |
| 0.7425714999844786, | |
| 0.6501901249866933 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 1.817464583000401, | |
| "cached_batched": 0.7273326250142418 | |
| }, | |
| "speedup": 2.4988080013113856 | |
| }, | |
| { | |
| "state": "video_speech", | |
| "prefix_tokens": 431, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 1.1875, | |
| "max_probability_delta": 0.18137001888040638, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 1.4375, | |
| "max_probability_delta": 0.20775812578824238, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 1.982235583011061, | |
| 2.0264377089915797, | |
| 2.0417945420194883 | |
| ], | |
| "cached_batched": [ | |
| 0.7379775000154041, | |
| 0.7445563329965807, | |
| 0.7704525000008289 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 2.0264377089915797, | |
| "cached_batched": 0.7445563329965807 | |
| }, | |
| "speedup": 2.721671442691074 | |
| }, | |
| { | |
| "state": "long_text", | |
| "prefix_tokens": 1250, | |
| "suffix_lengths": [ | |
| 111, | |
| 99, | |
| 111, | |
| 111 | |
| ], | |
| "cached_serial_vs_uncached": { | |
| "max_logit_delta": 1.28125, | |
| "max_probability_delta": 7.581746597207228e-08, | |
| "all_winners_match": true | |
| }, | |
| "cached_batch_vs_uncached": { | |
| "max_logit_delta": 1.390625, | |
| "max_probability_delta": 1.2115263459122616e-07, | |
| "all_winners_match": true | |
| }, | |
| "reversed_batch_vs_original": { | |
| "max_logit_delta": 0.0, | |
| "max_probability_delta": 0.0, | |
| "all_winners_match": true | |
| }, | |
| "prefix_unchanged": true, | |
| "branch_batch_sizes": [ | |
| 4 | |
| ], | |
| "warm_seconds": { | |
| "uncached": [ | |
| 1.2670293750124983, | |
| 1.2486699169967324, | |
| 1.2255972499842755 | |
| ], | |
| "cached_batched": [ | |
| 0.6019650840025861, | |
| 0.5762567500059959, | |
| 0.592385583993746 | |
| ] | |
| }, | |
| "warm_median_seconds": { | |
| "uncached": 1.2486699169967324, | |
| "cached_batched": 0.592385583993746 | |
| }, | |
| "speedup": 2.107866819746773 | |
| } | |
| ] | |
| } | |