E2B QAT gather tables (checkpoint-derived - pair with the QAT bundles)
Browse files- .gitattributes +5 -0
- ios-frontend/gemma4_qat_gather_raw/embed_per_layer.i8 +3 -0
- ios-frontend/gemma4_qat_gather_raw/embed_per_layer.scale.f32 +3 -0
- ios-frontend/gemma4_qat_gather_raw/embed_tokens.i8 +3 -0
- ios-frontend/gemma4_qat_gather_raw/embed_tokens.scale.f32 +3 -0
- ios-frontend/gemma4_qat_gather_raw/meta.json +12 -0
- ios-frontend/gemma4_qat_gather_raw/proj.f32 +3 -0
- ios-frontend/gemma4_qat_gather_raw/proj_norm.f32 +0 -0
.gitattributes
CHANGED
|
@@ -60,3 +60,8 @@ gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl/tokenizer/tokenizer.json filter=
|
|
| 60 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/original_model_0.mpsgraph filter=lfs diff=lfs merge=lfs -text
|
| 61 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/specialized_model_1.mpsgraph filter=lfs diff=lfs merge=lfs -text
|
| 62 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/original_model_0.mpsgraph filter=lfs diff=lfs merge=lfs -text
|
| 61 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/gemma4_e2b_qat_decode_int4lin_tbl.h18p.aimodelc/main-h18p-delegates/MPSGraph/mpsExecutable.mpsgraphpackage/specialized_model_1.mpsgraph filter=lfs diff=lfs merge=lfs -text
|
| 62 |
gpu-pipelined/gemma4_e2b_qat_decode_int4lin_tbl_aotc_h18p/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
| 63 |
+
ios-frontend/gemma4_qat_gather_raw/embed_per_layer.i8 filter=lfs diff=lfs merge=lfs -text
|
| 64 |
+
ios-frontend/gemma4_qat_gather_raw/embed_per_layer.scale.f32 filter=lfs diff=lfs merge=lfs -text
|
| 65 |
+
ios-frontend/gemma4_qat_gather_raw/embed_tokens.i8 filter=lfs diff=lfs merge=lfs -text
|
| 66 |
+
ios-frontend/gemma4_qat_gather_raw/embed_tokens.scale.f32 filter=lfs diff=lfs merge=lfs -text
|
| 67 |
+
ios-frontend/gemma4_qat_gather_raw/proj.f32 filter=lfs diff=lfs merge=lfs -text
|
ios-frontend/gemma4_qat_gather_raw/embed_per_layer.i8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf8818a39f6652df4f20a3d9aa8646cb65a11cb57390738894229f72ee42d76c
|
| 3 |
+
size 2348810240
|
ios-frontend/gemma4_qat_gather_raw/embed_per_layer.scale.f32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f85b1364ba6646b932817b30695a777a51c43a14a321ab7dfd8334a2cddfe789
|
| 3 |
+
size 1048576
|
ios-frontend/gemma4_qat_gather_raw/embed_tokens.i8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0998515764140526377de4d3f9a47819ccf8db29c866e455b45dffc6aaec0aa4
|
| 3 |
+
size 402653184
|
ios-frontend/gemma4_qat_gather_raw/embed_tokens.scale.f32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e29d1a6d2d0f5911028cc1c5f1d7aee292de0a11f7c0e0245f9db4418ffefafc
|
| 3 |
+
size 1048576
|
ios-frontend/gemma4_qat_gather_raw/meta.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"V": 262144,
|
| 3 |
+
"D": 1536,
|
| 4 |
+
"PLD": 8960,
|
| 5 |
+
"L": 35,
|
| 6 |
+
"ld": 256,
|
| 7 |
+
"embed_scale_main": 39.191835884530846,
|
| 8 |
+
"embed_scale_pl": 16.0,
|
| 9 |
+
"proj_scale": 0.02551551815399144,
|
| 10 |
+
"input_scale": 0.7071067811865476,
|
| 11 |
+
"rms_eps": 1e-06
|
| 12 |
+
}
|
ios-frontend/gemma4_qat_gather_raw/proj.f32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36a9c49bc7b1ee2977a9c25ce243f3284a6e4b6ad041841cd0d8292578f130be
|
| 3 |
+
size 55050240
|
ios-frontend/gemma4_qat_gather_raw/proj_norm.f32
ADDED
|
Binary file (1.02 kB). View file
|
|
|