JohnP1
/

d1a-e4b-mlx-q8

Text Classification

8-bit precision

Model card Files Files and versions

Commit History

Oct 25 7 days ago

13:36
config.json: vision_config

JohnP1

Verified signature
91f2c0f
06:57
Model card: wording

JohnP1

Verified signature
4092d49

Oct 17 15 days ago

13:58
Card: drop the note on where D1A is served

JohnP1

Verified signature
8d77f10

Oct 13 19 days ago

09:51
v0.6.1: routing temperature (use_case_temperatures.routing), same weights as v0.6 (jonpol01/d1a#264)

JohnP1

Verified signature
99d097a

Oct 12 20 days ago

10:49
Card: the chosen D1A avatar

JohnP1

Verified signature
028099c
10:47
Card: add the D1A avatar

JohnP1

Verified signature
927ac09

Oct 11 21 days ago

15:31
Card: say what D1A is first, with an example

JohnP1

Verified signature
7aac0b0

Oct 10 22 days ago

03:49
D1A-E4B v0.6

JohnP1

Verified signature
dbc89ea

Oct 7 25 days ago

01:21
D1A-E4B v0.5 MLX 8-bit (from JohnP1/d1a-e4b@v0.5): same build as v0.4, only the first weight shard differs

JohnP1

Verified signature
841021b

Oct 3 29 days ago

08:29
v0.4: pull-request labeling round 2, MLX 8-bit with media encoders

JohnP1

Verified signature
be470cb

Oct 2 30 days ago

23:54
v0.3: add the base's vision and audio encoders (media/) for photos and voice

JohnP1

Verified signature
583327b
23:30
v0.3: pull-request labeling (English and Japanese), MLX 8-bit

JohnP1

Verified signature
dd62a5e
08:20
D1A-E4B v0.2 (hybrid) MLX 8-bit, per-layer embeddings 4-bit; golden parity 7/274 flips, mean dp 0.020

JohnP1

Verified signature
8dffbc4

Oct 1

15:05
Calibration temperature 1.46 (fitted on the study's calibration split; the published head.pt had been left at 1.0)

JohnP1

Verified signature
c8eebe0
11:55
Card: parity numbers (3/274 changed answers)

JohnP1

Verified signature
cfb019f
09:40
D1A-E4B MLX q8 (per-layer embeddings 4-bit); parity check pending

JohnP1

Verified signature
7f53328
09:39
initial commit

JohnP1

Verified signature
4e06c71