Ben Graville commited on
README: ctx16384 bundles
Browse files
README.md
CHANGED
|
@@ -38,3 +38,14 @@ Same weights, same graph, same recipe as the 4096 bundles above — the exporter
|
|
| 38 |
| Bundle | Source checkpoint | Context | Functions | KV dtype | Producer | Created |
|
| 39 |
|---|---|---|---|---|---|---|
|
| 40 |
| `ctx8192/gpu-pipelined/gemma4_e4b_qat_decode_int4lin_tbl_pf64` | google/gemma-4-E4B-it-qat-q4_0-unquantized | 8192 | main+prefill | fp16 | coreai-core 1.0.0b2 | 20260817T204459Z |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
| Bundle | Source checkpoint | Context | Functions | KV dtype | Producer | Created |
|
| 39 |
|---|---|---|---|---|---|---|
|
| 40 |
| `ctx8192/gpu-pipelined/gemma4_e4b_qat_decode_int4lin_tbl_pf64` | google/gemma-4-E4B-it-qat-q4_0-unquantized | 8192 | main+prefill | fp16 | coreai-core 1.0.0b2 | 20260817T204459Z |
|
| 41 |
+
|
| 42 |
+
|
| 43 |
+
## High-context bundles — ctx 16384
|
| 44 |
+
|
| 45 |
+
Same weights, same graph, same recipe as the 4096 bundles above — the exporter's
|
| 46 |
+
`--max-ctx` only widens the dynamic sequence bound and the manifest's
|
| 47 |
+
`language.max_context_length`. KV cache stays fp16.
|
| 48 |
+
|
| 49 |
+
| Bundle | Source checkpoint | Context | Functions | KV dtype | Producer | Created |
|
| 50 |
+
|---|---|---|---|---|---|---|
|
| 51 |
+
| `ctx16384/gpu-pipelined/gemma4_e4b_qat_decode_int4lin_tbl_pf64` | google/gemma-4-E4B-it-qat-q4_0-unquantized | 16384 | main+prefill | fp16 | coreai-core 1.0.0b2 | 20260817T204711Z |
|