Datasets:
update nvidia-nemotron-cascade-2-30b-a3b-q4-k-m report with extended context data
Browse files
reports/nvidia-nemotron-cascade-2-30b-a3b-q4-k-m.md
CHANGED
|
@@ -69,6 +69,8 @@ Measured with `llama-bench`. All layers GPU-offloaded (`-ngl 99`).
|
|
| 69 |
| 4096 | 10,029 | 25.0 |
|
| 70 |
| 8192 | 9,508 | 20.6 |
|
| 71 |
| 16384 | 8,580 | 9.8 |
|
|
|
|
|
|
|
| 72 |
|
| 73 |
### Generation (tokens/s)
|
| 74 |
|
|
|
|
| 69 |
| 4096 | 10,029 | 25.0 |
|
| 70 |
| 8192 | 9,508 | 20.6 |
|
| 71 |
| 16384 | 8,580 | 9.8 |
|
| 72 |
+
| 32768 | 6,968 | 12.5 |
|
| 73 |
+
| 65536 | 4,967 | 7.0 |
|
| 74 |
|
| 75 |
### Generation (tokens/s)
|
| 76 |
|