card: Hub revision of the bundle upload; SHA256SUMS: README line
Browse files
README.md
CHANGED
|
@@ -390,7 +390,7 @@ table; an image adds one tower.
|
|
| 390 |
| `host/lm_head_fp16.bin` | the untied `lm_head.weight`, `[248320, 4096]` fp16, row-major, no header | 2,034,237,440 | `bfccf00b…1ba36d` |
|
| 391 |
| `host/lm_head_fp16.json` | its shape, dtype, source tensor and read-back check | 2,939 | `ef89e2fb…b75e86a` |
|
| 392 |
|
| 393 |
-
`SHA256SUMS` in the repository lists every file.
|
| 394 |
|
| 395 |
No AOT asset ships: the Swift runtime specializes the `.aimodel` files correctly on the Mac (the JIT
|
| 396 |
rows above); the decoder's AOT compile is 29.7 GB (fp16) / 25.7 GB (int8mix). The author's
|
|
|
|
| 390 |
| `host/lm_head_fp16.bin` | the untied `lm_head.weight`, `[248320, 4096]` fp16, row-major, no header | 2,034,237,440 | `bfccf00b…1ba36d` |
|
| 391 |
| `host/lm_head_fp16.json` | its shape, dtype, source tensor and read-back check | 2,939 | `ef89e2fb…b75e86a` |
|
| 392 |
|
| 393 |
+
`SHA256SUMS` in the repository lists every file. Uploaded 2026-10-03 as Hub revision `3d9e0ae355a52fc98b9e819b80c9c996734dc840` (the bundle upload); after the upload every file's size and sha256 matched the staged copy (the 8 LFS files by the Hub's own hash, the 23 small files by download). A later commit replaced only `README.md` and `SHA256SUMS`.
|
| 394 |
|
| 395 |
No AOT asset ships: the Swift runtime specializes the `.aimodel` files correctly on the Mac (the JIT
|
| 396 |
rows above); the decoder's AOT compile is 29.7 GB (fp16) / 25.7 GB (int8mix). The author's
|