jgalego commited on
Commit
4ce1f13
·
verified ·
1 Parent(s): 948bc7a

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +0 -4
README.md CHANGED
@@ -31,10 +31,6 @@ This is the standalone half of [Olmputer](https://huggingface.co/jgalego/olmpute
31
  | Parameters | 3.13B, almost all zero |
32
  | Vocabulary | the trace tokens (values in 15-bit chunks, jumps, stores, putchar, ...) plus PAD |
33
 
34
- Each generated token runs once through the 11 layers, and each layer does one part of executing the instruction:
35
-
36
- ![What each chip layer computes for one trace token](assets/diagram-layers.png)
37
-
38
  ![Heads per chip layer, by the memory they hold](assets/layers.png)
39
 
40
  Qwen3.5-4B's heads are 128 × 128, two value heads per key head; OLMo's are 96 × 192, one each. The changes that keep every read exact:
 
31
  | Parameters | 3.13B, almost all zero |
32
  | Vocabulary | the trace tokens (values in 15-bit chunks, jumps, stores, putchar, ...) plus PAD |
33
 
 
 
 
 
34
  ![Heads per chip layer, by the memory they hold](assets/layers.png)
35
 
36
  Qwen3.5-4B's heads are 128 × 128, two value heads per key head; OLMo's are 96 × 192, one each. The changes that keep every read exact: