README: today's E2B MLX build is v0.2 (v0.2.1 weights plus media)
Browse files
README.md
CHANGED
|
@@ -26,11 +26,12 @@ Use it where the GPU should stay free or power matters, not for speed.
|
|
| 26 |
## D1A builds compared
|
| 27 |
|
| 28 |
v0.6 is the current D1A. E2B v0.6 is not trained yet (jonpol01/d1a#229), so the E2B columns show the v0.6 plan and
|
| 29 |
-
today's E2B build, v0.2.1.
|
|
|
|
| 30 |
|
| 31 |
| | E2B · MLX | E2B · Core ML | E4B · MLX | E4B · Core ML |
|
| 32 |
|---|---|---|---|---|
|
| 33 |
-
| Status | v0.6 planned; v0.2
|
| 34 |
| Repository | JohnP1/d1a-e2b-mlx-q8 (public) | JohnP1/d1a-e2b-coreml (private) | JohnP1/d1a-e4b-mlx-q8 (public) | JohnP1/d1a-e4b-coreml (private) |
|
| 35 |
| Runs on | Mac GPU | Neural Engine: iPhone, iPad, Mac | Mac GPU | Neural Engine (tested on Macs only) |
|
| 36 |
| Text | yes | yes | yes | yes |
|
|
@@ -44,7 +45,7 @@ today's E2B build, v0.2.1.
|
|
| 44 |
| API | System One API (`d1a.serving.serve`), with `/v1/systemone/media` | Python reference only; no iOS app yet | System One API, served on the Mac mini | Python reference only (one process per chunk) |
|
| 45 |
| Size | 3.8 GB + 1.0 GB media | not built | 6.6 GB + 1.0 GB media | 6.1 GB (3.9 GB chunks + 2.1 GB embeddings) |
|
| 46 |
| Text speed (median) | ~0.4 s per 6-question request, M1 Max | pilot: ~0.95 s at 512 tokens, M4 | 0.63 s M4, 0.8 s M1 Max ¹ | 1.8 s M4, 2.9 s M1 Max ¹; load 52 s / 106 s |
|
| 47 |
-
| Media speed (median, M1 Max,
|
| 48 |
| Answer changes vs full precision | 5 of 274 (v0.2, vs PyTorch bf16) | pilot: 0 clear-cut | 0 of 38 clear-cut ¹ | 0 of 38 clear-cut ¹ |
|
| 49 |
|
| 50 |
¹ The same 28 real requests (42 questions, all ≤ 512 tokens), against the fp32 PyTorch build of v0.6. "Clear-cut"
|
|
|
|
| 26 |
## D1A builds compared
|
| 27 |
|
| 28 |
v0.6 is the current D1A. E2B v0.6 is not trained yet (jonpol01/d1a#229), so the E2B columns show the v0.6 plan and
|
| 29 |
+
today's E2B build: `v0.2`, the calibrated v0.2.1 weights with the media encoders (the `v0.2.1-2epoch-calibrated`
|
| 30 |
+
tag has the same weights without them).
|
| 31 |
|
| 32 |
| | E2B · MLX | E2B · Core ML | E4B · MLX | E4B · Core ML |
|
| 33 |
|---|---|---|---|---|
|
| 34 |
+
| Status | v0.6 planned; `v0.2` released | v0.6 planned; a v0.2.1 text pilot passed, not uploaded | **v0.6 released** | **v0.6 uploaded** (private) |
|
| 35 |
| Repository | JohnP1/d1a-e2b-mlx-q8 (public) | JohnP1/d1a-e2b-coreml (private) | JohnP1/d1a-e4b-mlx-q8 (public) | JohnP1/d1a-e4b-coreml (private) |
|
| 36 |
| Runs on | Mac GPU | Neural Engine: iPhone, iPad, Mac | Mac GPU | Neural Engine (tested on Macs only) |
|
| 37 |
| Text | yes | yes | yes | yes |
|
|
|
|
| 45 |
| API | System One API (`d1a.serving.serve`), with `/v1/systemone/media` | Python reference only; no iOS app yet | System One API, served on the Mac mini | Python reference only (one process per chunk) |
|
| 46 |
| Size | 3.8 GB + 1.0 GB media | not built | 6.6 GB + 1.0 GB media | 6.1 GB (3.9 GB chunks + 2.1 GB embeddings) |
|
| 47 |
| Text speed (median) | ~0.4 s per 6-question request, M1 Max | pilot: ~0.95 s at 512 tokens, M4 | 0.63 s M4, 0.8 s M1 Max ¹ | 1.8 s M4, 2.9 s M1 Max ¹; load 52 s / 106 s |
|
| 48 |
+
| Media speed (median, M1 Max, today's E2B) | photo 0.57 s, voice 0.19 s, video 1.94 s | not built | not measured | not built |
|
| 49 |
| Answer changes vs full precision | 5 of 274 (v0.2, vs PyTorch bf16) | pilot: 0 clear-cut | 0 of 38 clear-cut ¹ | 0 of 38 clear-cut ¹ |
|
| 50 |
|
| 51 |
¹ The same 28 real requests (42 questions, all ≤ 512 tokens), against the fp32 PyTorch build of v0.6. "Clear-cut"
|