Spaces:

77ethers
/

CarbonAlpha-demo

Paused

App Files Files Community

CarbonAlpha-demo

234 kB

Ctrl+K

Ctrl+K

1 contributor

History: 51 commits

77ethers's picture

News input: bump glow intensity — peak inner shadow 26px @ 40%, outer halo 48px @ 16%, border to 65% sage. Focus state matches with 22px+44px halo.

6b6d332 verified about 1 month ago

__pycache__
Fix action normalization in walkthrough app about 1 month ago
portfolio_env
replace gradio with custom openenv walkthrough ui about 1 month ago
static
News input: bump glow intensity — peak inner shadow 26px @ 40%, outer halo 48px @ 16%, border to 65% sage. Focus state matches with 22px+44px halo. about 1 month ago
.dockerignore

33 Bytes
switch custom walkthrough space to docker about 1 month ago
.gitattributes

1.52 kB
initial commit about 1 month ago
Dockerfile

772 Bytes
switch custom walkthrough space to docker about 1 month ago
README.md

1.91 kB
Switch to GRPO Phase-1 adapter (grpo_qwen25_7b_adapter_phase1_100_v1, 100 steps, 5/5 holdout beats baseline, mean regret 0.106). MODEL_SUBFOLDER env var lets us A/B back to the SFT adapter without code change. UI tag + README updated. about 1 month ago
app.py

48.9 kB
Remove DeepSeek judge entirely (server constants/functions/SSE phases + frontend HTML/CSS/JS/state). Comparison panel keeps the 3-way GRPO/SFT/base model progression — without an external judge. about 1 month ago
requirements.txt

177 Bytes
DeepSeek as independent judge: extends /api/plan-stream with judge_* phases that call deepseek/deepseek-v4-pro via HF Inference Providers (with V3 fallbacks). Frontend Judge panel under the comparison row shows winner badge + 1-line verdict + 4 structured bullets (format / reasoning / allocation / carbon). about 1 month ago