Commit History
-
release(2.0.1): Ollama serving-docs note (verify-it-stops)
Tiara RodneyUnverified signatureNo weights/eval change. 'Use with Ollama' notes the registry base carries [INST] + stops (no base setup needed) and adds a verify-it-stops check. Mirrors EuroLLM-9B-Teletype 2.0.1; Mistral needed no fix.e3c142a -
Release 2.0.0
Tiara RodneyUnverified signatureb9337f2 -
release(2.0.0): base-template retrain on posix-sdc v2.0.0
Tiara RodneyUnverified signatureNew r=16 adapter trained AND served with Mistral-7B-Instruct-v0.2's exact chat template (apply_chat_template, no placeholder markers), on the gate-hardened posix-sdc v2.0.0 corpus. Fixes the prior render mismatch where placeholder- trained weights no-op'd in deployment. - weights + teletype-lora-f16.gguf regenerated; tokenizer carries the base template verbatim (test-drive needs no custom template) - bounded/greedy held-out eval: operate 1.0, clean 14/16 (base 0/16, verified 0.13); card + loss/outcomes figures regenerated to match - training log_history + eval log refresheda5e0ea6
-
docs: report the base control measured first-party (0/16)
Tiara RodneyUnverified signatureEval section is now a base-vs-adapter A/B on the identical harness (--adapter none, same 16 held-out scenarios): bare Mistral-7B is 0/16 clean and 0/16 terminate against the adapter's 9/16. Replaces the 0/16 that was cited from an earlier run. Notes the verified_rate tell (the base satisfies the checker on 7/16 but never types exit). CHANGELOG 1.0.3.565c901
-
Release 1.0.2
Tiara RodneyUnverified signatureThe adapter packaged for ollama: ships teletype-lora-f16.gguf (GGUF LoRA). No weights/eval change from 1.0.1.7a2ab73 -
docs: prepare 1.0.2 release (changelog, ollama README, model-card prose cleanup)
Tiara RodneyUnverified signatureac661be -
chore: add the f16 GGUF LoRA adapter for ollama serving
Tiara RodneyUnverified signatureteletype-lora-f16.gguf (27MB): the r=16 adapter converted via llama.cpp convert_lora_to_gguf, tracked in LFS. Completes the Modelfile's 'ADAPTER ./teletype-lora-f16.gguf' path so the adapter serves over mistral:7b-instruct-v0.2 in ollama without merging the base.d45b2ac
-
Merge branch 'develop'
Tiara RodneyUnverified signature500fc9b -
Merge branch 'feature/7'
Tiara RodneyUnverified signatureb664799 -
todo(7): done
Tiara RodneyUnverified signatureIntro blockquote + 1.0.1 changelog committed on feature/7 (step 4 work commit), then released.4a20086 -
docs(7): clarify the intro is not tool-calling; changelog 1.0.1
Tiara RodneyUnverified signaturedbb2dac -
todo(7): in-progress
Tiara RodneyUnverified signatureREADME intro carries the not-tool-calling blockquote and CHANGELOG has [1.0.1], committed on the issue branch (step 4), then released as v1.0.1.2bbc3a3 -
todo(7): open
Tiara RodneyUnverified signature50f9c01 -
Merge branch 'feature/6'
Tiara RodneyUnverified signaturef7d78de -
todo(6): done
Tiara RodneyUnverified signatureIntro blockquote added: 'not a tool-using model'; pushed to HF main.9a0d5cc -
todo(6): in-progress
Tiara RodneyUnverified signatureThe README intro carries a clear 'not a tool-using model' callout before the rest of the card.acd58e1 -
todo(6): open
Tiara RodneyUnverified signature58ba2d0 -
Merge branch 'develop'
Tiara RodneyUnverified signature79e7a00 -
Merge branch 'feature/5'
Tiara RodneyUnverified signaturede74f90 -
todo(5): done
Tiara RodneyUnverified signatureRenamed; layout, LFS, and CHANGELOG in place; gitflow + mime-todo bootstrapped.16b25cc -
todo(5): in-progress
Tiara RodneyUnverified signatureRepo is Mistral-7B-Teletype; adapter loads from the Hub id; safetensors via LFS; CHANGELOG + attributes/ignore present.54a3696 -
todo(5): open
Tiara RodneyUnverified signature9efa10f -
Merge branch 'feature/4'
Tiara RodneyUnverified signature3e36b82 -
todo(4): done
Tiara RodneyUnverified signatureThesis + Evaluation sections present and honest.7999bec -
todo(4): in-progress
Tiara RodneyUnverified signatureCard has a thesis section and an evaluation section with the metrics, the failure-mode reading, and the cited-and-flagged 0/16.b43f42b -
todo(4): open
Tiara RodneyUnverified signaturea7e3c69 -
Merge branch 'feature/3'
Tiara RodneyUnverified signaturef9cde2e -
todo(3): done
Tiara RodneyUnverified signaturefigures/ complete; sources committed and regenerable.625d8e1 -
todo(3): in-progress
Tiara RodneyUnverified signaturefigures/ holds 6 PNGs plus their .puml/.gp/.dat sources; the card embeds them.84d172c -
todo(3): open
Tiara RodneyUnverified signature1304022 -
Merge branch 'feature/2'
Tiara RodneyUnverified signature57d6564 -
todo(2): done
Tiara RodneyUnverified signatureTrained via sekft-train --hub; loss 6.5->0.1; card updated.3057637 -
todo(2): in-progress
Tiara RodneyUnverified signatureadapter_model.safetensors is the 787-trajectory fp16 run; the training table reflects 787 / fp16 / r=16 / 3 epochs.2b4272c -
todo(2): open
Tiara RodneyUnverified signature442a2fd -
Merge branch 'feature/1'
Tiara RodneyUnverified signaturef2b58db -
todo(1): done
Tiara RodneyUnverified signatureShipped as posix-sdc-mistral-v02-7b-lora, the v0.1 proof of concept.2cf0ac6 -
todo(1): in-progress
Tiara RodneyUnverified signatureThe prior adapter and card exist with the 0/16->9/16 result and an Ollama Modelfile.cf1d24a -
todo(1): open
Tiara RodneyUnverified signaturef6efba7 -
Mistral-7B-Teletype 1.0.0: self-directed shell-operation adapter for Mistral-7B
Tiara RodneyUnverified signatureLoRA adapter trained on posix-sdc v1.2.2 (787 trajectories, fp16 r=16), a research-grade model card with PlantUML/gnuplot figures, the held-out generalization result (operate_rate 1.0, 9/16 clean), CHANGELOG, and an Ollama Modelfile.c3e401b