Customise a model from the portal: a file, a model, a button

A page for anyone in a project, built for the one who knows least: an
Excel, CSV or JSONL file of questions and answers, previewed; a model; a
name; who may use it; and a button. The weights to train on are found
from where the model's came (a GGUF repository's base model), the card's
memory from their size, the node as the one with the most free. When it
is done the adapter is published by itself (a llama.cpp model's replicas
replaced to load it) and is a model of its own in the chat and the API.

The project's by default; its members share it wider (confirmed, and
never wider than the knowledge bases its pairs came from) or delete it.
One trains at a time per project. ctl train start needs only --model and
--data now, and --publish publishes when done.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Ilker ManapandClaude Opus 5.5 committed 2026-11-19 11:31:12 +01:00
1 parent fcda4d3f85
commit a53b2f4be9
33 files changed
+2199 -134

No files matched your search

+12 -5
View File
@@ -385,18 +385,25 @@ right and mixed facts up on the rest, while the same questions through the
knowledge base got 9 of 9. Use both: the knowledge base for what answers
say, an adapter for how they sound.
Anyone in a project can do it from the portal: **Customise a model**
(`/train`) takes an Excel, CSV or JSONL file of questions and answers, a
model, a name and who may use it, and publishes the adapter by itself when it
is trained. The weights, the node and the card's memory are worked out. From
the command line:
```sh
# 1. Questions and answers from the collection, written by a model of the
# cluster. Read the file through and correct it.
llmgw ctl kb dataset Acme/policies --model qwen2.5-coder-14b-instruct --out policies.jsonl
# 2. Train on a node. --gpu-memory is what the job needs of a card; it
# runs beside an engine when the engine leaves that much free.
llmgw ctl train start policies-tone --model qwen2.5-0.5b-instruct \
--base-repo Qwen/Qwen2.5-0.5B-Instruct --node cuda01 --gpu-memory 2300 --data policies.jsonl
# 2. Train. The weights to train on are found from where the model's came
# (--base-repo names others), the card's memory from their size
# (--gpu-memory), the node as the one whose card has the most free
# (--node); it runs beside an engine when the engine leaves that much.
llmgw ctl train start policies-tone --model qwen2.5-0.5b-instruct --data policies.jsonl
llmgw ctl train show JOB --follow
# 3. Make it the model's adapter, kept on the mirror.
# 3. Make it the model's adapter, kept on the mirror (or --publish in step 2).
llmgw ctl train publish JOB --restart
```