Christian Medina
|
50fa171f68
|
fix: use local model path instead of HuggingFace download
|
2026-06-30 20:24:44 -04:00 |
|
Christian Medina
|
7f7467ef89
|
fix: remove max_seq_length from TrainingArguments (passed to SFTTrainer)
|
2026-06-30 20:17:04 -04:00 |
|
Christian Medina
|
7181acc580
|
fix: save train/test.jsonl to output_dir not parent
|
2026-06-30 20:01:05 -04:00 |
|
Christian Medina
|
d5ba471b98
|
fix: auto-split dataset if split files missing
|
2026-06-30 19:54:50 -04:00 |
|
Christian Medina
|
bbdbe91871
|
fix: fix dataset path calculation
|
2026-06-30 19:01:39 -04:00 |
|
Christian Medina
|
14a302dc5e
|
fix: make scripts executable
|
2026-06-30 18:49:39 -04:00 |
|
Christian Medina
|
97004261f5
|
fix: multiple bug fixes - remove unused imports, fix typo, fix dataset paths, add ai command fallback
|
2026-06-30 18:48:46 -04:00 |
|
Christian Medina
|
75aeeeb20a
|
fix: use torchrun for distributed training, load model on CPU
|
2026-06-30 18:45:14 -04:00 |
|
Christian Medina
|
5ad5518ca6
|
fix: add deepspeed to auto-install dependencies
|
2026-06-30 18:30:23 -04:00 |
|
Christian Medina
|
4f2f8ef03f
|
feat: use DeepSpeed ZeRO-3 for optimal 2-GPU training
|
2026-06-30 18:26:48 -04:00 |
|
Christian Medina
|
a04bf4a9cf
|
feat: add run_with_accelerate.sh helper script
|
2026-06-30 18:26:08 -04:00 |
|
Christian Medina
|
cd58ffd31e
|
fix: use FSDP for 2-GPU training, reduce batch size
|
2026-06-30 18:23:34 -04:00 |
|
Christian Medina
|
3d87f0b8b3
|
fix: force reinstall PyTorch to resolve version conflicts
|
2026-06-30 17:58:52 -04:00 |
|
Christian Medina
|
ef42e67183
|
fix: use CompressedTensors for Ornith, add torch import, expand LoRA targets
|
2026-06-30 17:54:35 -04:00 |
|
Christian Medina
|
7e8ddabf03
|
fix: remove dataset download, it's already in repo
|
2026-06-30 17:11:29 -04:00 |
|
Christian Medina
|
28d0a956fd
|
fix: update PyTorch to cu124 for RTX 5090, fix total_memory typo
|
2026-06-30 16:58:11 -04:00 |
|
Christian Medina
|
1786830d84
|
feat: unload AI model before training
|
2026-06-30 16:00:17 -04:00 |
|
Christian Medina
|
2074d3d993
|
feat: use Ornith-1.0-35B-FP8 with multi-GPU support
|
2026-06-30 15:57:54 -04:00 |
|
Christian Medina
|
215886d6a7
|
fix: use $HOME/loras by default
|
2026-06-30 15:51:28 -04:00 |
|
Christian Medina
|
11951a8276
|
feat: add deploy-agenx-lora.sh to repo
|
2026-06-30 15:38:05 -04:00 |
|
Christian Medina
|
39a493f261
|
fix: remove deploy-agenx-lora.sh from repo
|
2026-06-30 15:35:42 -04:00 |
|
Christian Medina
|
11653ed743
|
feat: add simple deploy-agenx-lora.sh script
|
2026-06-30 15:31:13 -04:00 |
|
Christian Medina
|
44939e374b
|
refactor: simplify deploy script, add train-on-this-server.sh
|
2026-06-30 15:22:54 -04:00 |
|
Christian Medina
|
0614166c7a
|
fix: use remote Gitea directly, remove local fallback
|
2026-06-30 15:20:03 -04:00 |
|
Christian Medina
|
6946ebac1a
|
feat: use local Gitea when available, fallback to remote
|
2026-06-30 15:15:51 -04:00 |
|
Christian Medina
|
8e51c39c6e
|
fix: remove large dataset from git tracking, add to .gitignore
|
2026-06-30 15:06:13 -04:00 |
|
Christian Medina
|
e1b80301d8
|
refactor: update deploy script to match cyron pattern
|
2026-06-30 15:03:40 -04:00 |
|
Christian Medina
|
3d2ae812a7
|
feat: add deploy-and-train.sh script for server deployment
|
2026-06-30 15:02:13 -04:00 |
|
Christian Medina
|
418a4cc76d
|
init: add LoRA training infrastructure and 20k dataset
|
2026-06-30 14:49:44 -04:00 |
|