Nov 2 2 days ago

1. 21:47

   new model repo name + full MATH-500 numbers in header Running

   akhaliq HF Staff
   Verified signature

   56e478e
2. 21:20

   stream tokens live via TextIteratorStreamer

   akhaliq HF Staff
   Verified signature

   f26d031
3. 21:15

   fix: pre-warm torch.linalg CUDA lazy loader before generation threads (#90613)

   akhaliq HF Staff
   Verified signature

   435bf7c
4. 20:57

   fix: unpack BatchEncoding before generate (input_ids/attention_mask)

   akhaliq HF Staff
   Verified signature

   41d30d6
5. 20:54

   graceful fallback when no GPU attached

   akhaliq HF Staff
   Verified signature

   50fcd50
6. 20:52

   demo app: live base-vs-edited compare + recorded runs

   akhaliq HF Staff
   Verified signature

   4a92b64