@seantalts/stanli 0.9.2 → 0.9.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +6 -5
- package/package.json +1 -1
- package/stanli-compiler.js +6320 -6314
- package/stanli.wasm +0 -0
package/README.md
CHANGED
|
@@ -46,11 +46,12 @@ can precompile it at build time (`stanc --O1 --debug-optimized-mir model.stan`)
|
|
|
46
46
|
and pass `mir` instead of `code`; neither browser compiler loads, and the WASM
|
|
47
47
|
runtime alone is ~1.5 MB gzipped.
|
|
48
48
|
|
|
49
|
-
On Eight Schools, portable MIR is
|
|
50
|
-
33,320 bytes (2,000 gzipped) for legacy MIR.
|
|
51
|
-
build, median decoder parsing was
|
|
52
|
-
preparation was
|
|
53
|
-
measurements; source compilation still dominates that
|
|
49
|
+
On Eight Schools, compact portable MIR is 6,932 bytes (1,793 gzipped), compared
|
|
50
|
+
with 33,320 bytes (2,000 gzipped) for legacy MIR. Across 51 fresh processes on
|
|
51
|
+
the same Apple arm64 release build, median decoder parsing was 0.074 ms versus
|
|
52
|
+
0.293 ms, and complete preparation was 0.278 ms versus 0.682 ms. These are
|
|
53
|
+
one-time preparation measurements; source compilation still dominates that
|
|
54
|
+
path.
|
|
54
55
|
|
|
55
56
|
118 of 119 posteriordb corpus models verify against CmdStan's log
|
|
56
57
|
density, gradients, and write_array values from inside this WASM build;
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@seantalts/stanli",
|
|
3
|
-
"version": "0.9.
|
|
3
|
+
"version": "0.9.3",
|
|
4
4
|
"description": "Full Stan in the browser: stanc3 compiles the model in JS, a WASM runtime lowers it to an op graph and runs NUTS. No server, no C++ toolchain.",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"main": "index.mjs",
|