Strip heatmap leak from prompt; let model discover peak hours via tools

Upload folder using huggingface_hub

Merge HF run-output upload

Inject peak hours + history + post-mandate, run SFT every round

Upload folder using huggingface_hub

ReAct two-pass per day so model sees current-day tool results

Strip leaked peak-hour info from observation, force tool discovery

Mandate tool calls in system prompt to debug zero-tool collapse

Upload folder using huggingface_hub

Match eval sampling to training, log all I/O, single round

Upload folder using huggingface_hub

train(grpo): unified hint prompt, no-history chat, positive-advantage filter

Upload folder using huggingface_hub

fix: align notebook with 15-day horizon, drop unused replies field

update

Merge branch 'main' of https://huggingface.co/spaces/vaibhavkhandare/train-bhai-train

train: batched parallel rollouts on Qwen2.5-3B + parser hardening

Stop tracking plots/*.png with Git LFS; use small inline PNGs for HF Hub.

Default repo clone branch to main for training notebooks and HF script.

Set TASK_HORIZON to 15 days and align graders, UI, and training prompts.

update

update

Upload folder using huggingface_hub

fix(env): tolerate malformed predict_engagement scheduled_actions

train: shrink to weekly horizon + bounded steps

train: default HF Job flavor l4x1 -> l40sx1 (48GB VRAM)

train: per-step credit + drop replies + larger batches

fix(notebook): py3.11 f-string backslash error in format_obs

Track PNGs with LFS

Merge branch 'main' of https://huggingface.co/spaces/vaibhavkhandare/train-bhai-train

update

fix(notebook): pin typing_extensions>=4.13.0 to fix pydantic Sentinel ImportError

fix: add missing metadata on display_data outputs

fix: add missing 'name' to stream outputs

add: viraltest code (server, models, inference)

add: train_grpo notebook

initial commit

fix: restore parse_model_output exception parity with original bare except

chore: align train_grpo.ipynb with smoke/syntax patterns for Colab

add training/syntax_only.ipynb — kernel + Python syntax only (no project logic)

add train_grpo_smoke notebook; quote pip versions in train_grpo

fix: notebook loads Qwen without bitsandbytes on Mac; optional training deps

fix: robust notebook setup (no magic shell) + local CWD auto-detect

update

Merge branch 'hack1' of github.com:VaibhavKhandare/viral-posts-env into hack1

update

fix: rewrite training notebook for real LoRA fine-tuning on Colab

la la la --123

firstiteration

reduced steps to fit out free tier