forge: make your own foundation model -- new, Minitron-style prune, p… - #136
Merged
Merged
Conversation
…retrain / distill, lineage card pollard-forge new builds a fresh qwen3 / llama / qwen2 architecture (size presets or exact dims; tokenizer copied or trained). prune carves a smaller model out of a teacher: drops the layers with the lowest block influence and the FFN channels that fire least, measured on --calib. train pretrains, continues or distills (CE + temperature-scaled KL), mixes text / jsonl / hf: sources by weight, checkpoints and resumes, runs under torchrun. card writes the lineage from forge.json and requires a license. Output is a normal HF checkpoint, so pollard --hf <dir> --run quantizes it like any model. Checked end to end: SmolLM2-135M -> 82M prune -> distill (held-out loss 5.62 -> 3.86 in 30 steps) -> card -> GGUF -> llama-cli generates.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…retrain / distill, lineage card
pollard-forge new builds a fresh qwen3 / llama / qwen2 architecture (size presets or exact dims; tokenizer copied or trained). prune carves a smaller model out of a teacher: drops the layers with the lowest block influence and the FFN channels that fire least, measured on --calib. train pretrains, continues or distills (CE + temperature-scaled KL), mixes text / jsonl / hf: sources by weight, checkpoints and resumes, runs under torchrun. card writes the lineage from forge.json and requires a license. Output is a normal HF checkpoint, so pollard --hf
--run quantizes it like any model. Checked end to end: SmolLM2-135M -> 82M prune -> distill (held-out loss 5.62 -> 3.86 in 30 steps) -> card -> GGUF -> llama-cli generates.