renaudb1999/le-harnais-ft-counsel-Meta-Llama-3.1-8B-Instruct-jepa-loraembed

le-harnais / ft-counsel-Meta-Llama-3.1-8B-Instruct-jepa-loraembed

⚠️ retrain-only — this is an ablation checkpoint, not useful for inference. It ships only to reproduce / continue the training study. For real use see the hero models: le-harnais-ft-agentworld-{1b,3b,8b}, le-harnais-ft-counsel.

Counsel-corpus ablation (scaling / data-augmentation / JEPA). JEPA ≈ +7 @3B, ≈0 @8B.

• —Base model: `meta-llama/Meta-Llama-3.1-8B-Instruct` — Built with Llama; Llama Community License applies.
• —Class: ablation
• —Training data: datasets/counsel_train.jsonl (270 ex; wisdom+commentary, PD sources)
• —Headline: counsel-corpus scaling × augmentation × JEPA ablation (see docs/jepa.md)

Reproduce

see docs/jepa.md (counsel scaling grid)

Full recipe, datasets, and eval commands: see docs/REPRODUCE.md in the [le-harnais distribution]. Provenance & license: docs/PROVENANCE.md.

Formats in this repo

• —*.safetensors — bf16 inference weights (serve with transformers or le-harnais lh-serve/candle).
• —*.Q4_K_M.gguf — portable 4-bit quant (run via ollama / llama.cpp; Mac-friendly).

Orchestration amplifies a capable generator; it does not create competence.