[gemini] [HARNESS] Switch model default from Anthropic cloud to local Ollama (sovereignty) (#22 )

v1.0 rejected — NaN from wrong tokenizer, Morrowind MCP pipeline working
2026-03-26 16:40:54 -04:00 · 2026-03-26 12:32:08 -04:00
1 changed files with 4 additions and 3 deletions
--- a/autolora/manifest.yaml
+++ b/autolora/manifest.yaml
@@ -55,7 +55,8 @@ adapters:
  timmy-v1.0:
    base: hermes4-14b-4bit
    date: 2026-03-26
-    status: training
-    data: 1125 train / 126 valid (same curated set, reused)
+    status: rejected
+    data: 1125 train / 126 valid (same curated set, reused from 8B — NOT re-tokenized)
    training: { lr: 1e-6, rank: 16, iters: 800 }
-    notes: "First 14B adapter. Conservative lr for new arch."
+    eval: "Val NaN iter 100, train NaN iter 160. Dead."
+    notes: "Data was pre-truncated for Llama3 tokenizer, not Qwen3. Must re-run clean_data.py with 14B tokenizer before v1.1."
Author	SHA1	Message	Date
Alexander Whitestone	3bcf8168e7	[gemini] [HARNESS] Switch model default from Anthropic cloud to local Ollama (sovereignty) (#22 )	2026-03-26 16:40:54 -04:00
Alexander Whitestone	f9155b28e3	v1.0 rejected — NaN from wrong tokenizer, Morrowind MCP pipeline working	2026-03-26 12:32:08 -04:00