aleleba
0 Followers · 0 Following
  • Joined on 2024-12-27
Repositories
30
 Projects Packages Public Activity Starred Repositories
aleleba created pull request aleleba/qwen3-6-lora#3 2026-07-29 13:02:43 -06:00
Phase 4: merge LoRA adapter and run full evaluation gates on the merged checkpoint
aleleba pushed to agente-fase4-merge-eval at aleleba/qwen3-6-lora 2026-07-29 13:02:16 -06:00
19dc5f3227 Fase 4: resultados de las puertas 2, 3 y 4 sobre el checkpoint mergeado
2762761a8e Fase 4: puerta 1 - reportar tambien loss ponderado por token (comparable a Trainer)
ceba5f80cd Fase 4: puerta 1 (eval-loss offline por bucket) y contenedor/scripts de puertas 2-4
268451aed7 Fase 4: scripts/20_merge_lora.py - merge streaming shard-a-shard del LoRA sobre el checkpoint base
Compare 4 commits »
aleleba created branch agente-fase4-merge-eval in aleleba/qwen3-6-lora 2026-07-29 13:02:13 -06:00
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-29 10:16:55 -06:00
7e57987385 Fase 3: versionar el adapter LoRA final (pesos + tokenizer + config)
aleleba deleted branch agente-fase3-training from aleleba/qwen3-6-lora 2026-07-29 07:13:50 -06:00
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-29 07:13:48 -06:00
c275b2ffbf Merge pull request 'agente-fase3-training: LoRA training script and completed training run' (#2) from agente-fase3-training into master
c4a0c404aa Fase 3: scripts/10_train.py - entrenamiento LoRA con masking manual, target_modules confirmados (incluye out_proj de Gated DeltaNet)
Compare 2 commits »
aleleba merged pull request aleleba/qwen3-6-lora#2 2026-07-29 07:13:48 -06:00
agente-fase3-training: LoRA training script and completed training run
aleleba created pull request aleleba/qwen3-6-lora#2 2026-07-29 04:40:34 -06:00
agente-fase3-training: LoRA training script and completed training run
aleleba pushed to agente-fase3-training at aleleba/qwen3-6-lora 2026-07-29 04:40:11 -06:00
c4a0c404aa Fase 3: scripts/10_train.py - entrenamiento LoRA con masking manual, target_modules confirmados (incluye out_proj de Gated DeltaNet)
aleleba created branch agente-fase3-training in aleleba/qwen3-6-lora 2026-07-29 04:40:10 -06:00
aleleba deleted branch agente-fase2-dataset from aleleba/qwen3-6-lora 2026-07-28 19:54:50 -06:00
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-28 19:54:44 -06:00
5f912a15db Merge pull request 'agente-fase2-dataset: Build Phase 2 v1 training dataset from sanitized sources and synthetic seeds' (#1) from agente-fase2-dataset into master
3f14aa5cec Fase 2: 06_validate_dataset.py - validacion con tokenizer/chat_template real (1470/1470 ok, 0 excepciones, 0 filtrados, 0 secretos)
ab5d78b5c8 Fase 2: 05_build_dataset.py - ensamblado v1 (1470 ejemplos: 750 nuevos + 720 replay, split 90/10 estratificado por bucket)
73e7f721cd Fase 2: seeds de los 6 buckets (294 ejemplos: penpot 41, otros_mcps 102, skills_adherencia 55, delegacion_subagentes 25, negativos 46, manejo_errores 25)
837f91060f Fase 2: 04_sanitize.py - scrubbing de secretos de skills/agentes/plans (10 secretos unicos, gate en verde)
Compare 5 commits »
aleleba merged pull request aleleba/qwen3-6-lora#1 2026-07-28 19:54:43 -06:00
agente-fase2-dataset: Build Phase 2 v1 training dataset from sanitized sources and synthetic seeds
aleleba created pull request aleleba/qwen3-6-lora#1 2026-07-28 18:58:16 -06:00
agente-fase2-dataset: Build Phase 2 v1 training dataset from sanitized sources and synthetic seeds
aleleba pushed to agente-fase2-dataset at aleleba/qwen3-6-lora 2026-07-28 18:57:54 -06:00
3f14aa5cec Fase 2: 06_validate_dataset.py - validacion con tokenizer/chat_template real (1470/1470 ok, 0 excepciones, 0 filtrados, 0 secretos)
ab5d78b5c8 Fase 2: 05_build_dataset.py - ensamblado v1 (1470 ejemplos: 750 nuevos + 720 replay, split 90/10 estratificado por bucket)
73e7f721cd Fase 2: seeds de los 6 buckets (294 ejemplos: penpot 41, otros_mcps 102, skills_adherencia 55, delegacion_subagentes 25, negativos 46, manejo_errores 25)
837f91060f Fase 2: 04_sanitize.py - scrubbing de secretos de skills/agentes/plans (10 secretos unicos, gate en verde)
Compare 4 commits »
aleleba created branch agente-fase2-dataset in aleleba/qwen3-6-lora 2026-07-28 18:57:51 -06:00
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-28 06:52:45 -06:00
21bc219f31 Fase 1: dataset de replay generado (720 ejemplos, ok=564 truncated=156 failed=0)
ef35472398 Fix: OUTPUT_PATH de 03_build_replay.py apuntaba a /workspace/... (ruta de contenedor inexistente en el host)
4d73be8e1b gitignore: excluir .worktrees/ (usado por agentes autonomos en background)
Compare 3 commits »
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-27 23:35:06 -06:00
b2aa855a70 Fase 1: schemas de los 5 MCPs y script de generacion de replay
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-27 21:52:18 -06:00
d121c4465e Fase 0: scripts de verificacion de hardware e inspeccion de modulos
aleleba pushed to master at aleleba/qwen3-6-lora 2026-07-27 21:42:48 -06:00
572ee0b60e Fase 0: estructura base de carpetas y docker-compose de training