Anthracite

community

Activity Feed

AI & ML interests

None defined yet.

Recent Activity

intervitens new activity 12 days ago

anthracite-org/pixmo-point-explanations-images:[bot] Conversion to Parquet

lucyknada new activity 12 days ago

anthracite-org/magnum-v4-22b:It's really good.

intervitens updated a dataset 14 days ago

anthracite-org/pixmo-point-explanations-images

View all activity

anthracite-org's activity

grimjim

posted an update 3 days ago

Post

1245

A reminder that literal base models are valid choices for base model in task arithmetic mergers. Each Instruct or fine-tuned model then becomes a vector against the base model. Example merge formula used can be found via this model page.
grimjim/Magnolia-v3-12B

intervitens

in anthracite-org/pixmo-point-explanations-images 12 days ago

[bot] Conversion to Parquet

#1 opened 13 days ago by

parquet-converter

lucyknada

in anthracite-org/magnum-v4-22b 12 days ago

It's really good.

#4 opened 13 days ago by

FistfulSteel

intervitens

updated a dataset 14 days ago

anthracite-org/pixmo-point-explanations-images

Viewer • Updated 14 days ago • 71.7k • 19 • 1

intervitens

updated a dataset 19 days ago

anthracite-org/pixmo-cap-qa-images

Viewer • Updated 19 days ago • 269k • 152

intervitens

in anthracite-org/pixmo-cap-images 20 days ago

[bot] Conversion to Parquet

#1 opened 20 days ago by

parquet-converter

intervitens

updated a dataset 21 days ago

anthracite-org/pixmo-cap-images

Viewer • Updated 21 days ago • 708k • 124 • 1

lucyknada

in anthracite-org/magnum-v4-12b 22 days ago

What class does this model belong to?

#7 opened 22 days ago by

FrescoHF

Delta-Vector

in anthracite-org/magnum-v4-72b-exl2 22 days ago

2.75bpw?

#2 opened 22 days ago by

Darkknight535

grimjim

posted an update 24 days ago

Post

943

Speculative decoding only requires that the tokenizers for the two LLMs used line up; the model architectures do not have to be otherwise compatible. As proof of concept, I used exllamav2 to run Llama 3.2 1B Instruct (at 6bpw, for speed) as the draft model to accelerate the target model of a Llama 3 8B merge of Instruct models (at 8bpw, for accuracy). The difference between tokenizers was minor enough to allow this. With 8k context length allocated for each model, both fit in under 13GB VRAM.
https://github.com/turboderp/exllamav2
meta-llama/Llama-3.2-1B-Instruct
grimjim/llama-3-Nephilim-v3-8B

The proof-of-concept Python script compared a zero-shot creative task of writing a story limited to 500 tokens. Speculative decoding improved performance by approximately one third (e.g., increasing from 31 tokens/sec to 46 tokens/sec) over conventional decoding, and was consistent over a few runs. While not statistically significant, this implies that smaller models aimed at edge computing can serve effectively as draft models in the general case.

It is straightforward to consult literature to affirm that fine-tuning draft models can be a way of inducing behavioral change in target models, in a manner not unlike how samplers can be used to induce changes. I speculate that the impact of a fine-tuned draft model would be on part with a LoRA (Low-Rank Adaptation), as the target model retains veto power. The small size of draft model candidates means that more people can perform local full fine-tuning.

It is intuitively obvious that a distilled model can be used as a draft model for the larger teacher model so long as tokenizers line up; e.g., a distilled 8B model can draft for a 70B teacher model. Perhaps Llama-3.1-SuperNova-Lite 8B could effectively draft for the original Llama-3.1-405B-Instruct model.
arcee-ai/Llama-3.1-SuperNova-Lite
meta-llama/Llama-3.1-405B-Instruct