Hugging Face's logo

ngquocvinh
/
EmbeddingGemma-2-GGUF

Feature Extraction
GGUF
sentence-transformers
llama.cpp
embedding
multimodal
multilingual
quantized
imatrix
Model card Files
xet
 Community
ngquocvinh commited on
Commit
fd2fec2
·
verified ·
1 Parent(s): 893ef53

Add IQ3_S after text embedding smoke pass

Browse files
Files changed (5) hide show
  1. .gitattributes +1 -0
  2. EmbeddingGemma-2-IQ3_S.gguf +3 -0
  3. SHA256SUMS.txt +1 -0
  4. reproducibility/artifact-measurements.tsv +1 -0
  5. reproducibility/manifest.md +8 -0
.gitattributes CHANGED
@@ -51,3 +51,4 @@ EmbeddingGemma-2-Q2_K_S.gguf filter=lfs diff=lfs merge=lfs -text
51
  EmbeddingGemma-2-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
52
  EmbeddingGemma-2-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text
53
  EmbeddingGemma-2-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
 
 
51
  EmbeddingGemma-2-IQ4_XS.gguf filter=lfs diff=lfs merge=lfs -text
52
  EmbeddingGemma-2-IQ4_NL.gguf filter=lfs diff=lfs merge=lfs -text
53
  EmbeddingGemma-2-IQ3_M.gguf filter=lfs diff=lfs merge=lfs -text
54
+ EmbeddingGemma-2-IQ3_S.gguf filter=lfs diff=lfs merge=lfs -text
EmbeddingGemma-2-IQ3_S.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce15d1c47c0b9bd3ab51921b8fbf4758b767fe8ae035a81b44783e37d01a652d
3
+ size 142545536
SHA256SUMS.txt CHANGED
@@ -1,6 +1,7 @@
1
  a7788cc0daf653026d9762fc90e351763e472427cb2ac9eef1cdb3d89538057a EmbeddingGemma-2-IQ1_M.gguf
2
  c4e59a215a94e4ab077e1a10c43df85ab1c63addc8fdf5515a5ee9b99c8a15f4 EmbeddingGemma-2-IQ2_XS.gguf
3
  173b0e1daabbd79873c5757dc0d44cefb269927286b978f4685f5232831ac90a EmbeddingGemma-2-IQ3_M.gguf
 
4
  1ae7f62c21191734a7c8c4f2708537e29e5ca99bfe354284551ba8249f975e80 EmbeddingGemma-2-IQ4_NL.gguf
5
  96db13a1041a7ef09bf0ca3a7e582b3331dae326a3aca9898a93f257979af427 EmbeddingGemma-2-IQ4_XS.gguf
6
  4953a59ae0d8a725052407f458101ef81d8acb6a95f3293f4949ff3f594ef18a EmbeddingGemma-2-Q1_0.gguf
 
1
  a7788cc0daf653026d9762fc90e351763e472427cb2ac9eef1cdb3d89538057a EmbeddingGemma-2-IQ1_M.gguf
2
  c4e59a215a94e4ab077e1a10c43df85ab1c63addc8fdf5515a5ee9b99c8a15f4 EmbeddingGemma-2-IQ2_XS.gguf
3
  173b0e1daabbd79873c5757dc0d44cefb269927286b978f4685f5232831ac90a EmbeddingGemma-2-IQ3_M.gguf
4
+ ce15d1c47c0b9bd3ab51921b8fbf4758b767fe8ae035a81b44783e37d01a652d EmbeddingGemma-2-IQ3_S.gguf
5
  1ae7f62c21191734a7c8c4f2708537e29e5ca99bfe354284551ba8249f975e80 EmbeddingGemma-2-IQ4_NL.gguf
6
  96db13a1041a7ef09bf0ca3a7e582b3331dae326a3aca9898a93f257979af427 EmbeddingGemma-2-IQ4_XS.gguf
7
  4953a59ae0d8a725052407f458101ef81d8acb6a95f3293f4949ff3f594ef18a EmbeddingGemma-2-Q1_0.gguf
reproducibility/artifact-measurements.tsv CHANGED
@@ -17,3 +17,4 @@ EmbeddingGemma-2-Q2_K_S.gguf 116445824 0.116446 PASS_EMBEDDING_CPU
17
  EmbeddingGemma-2-IQ4_XS.gguf 169382528 0.169383 PASS_EMBEDDING_CPU
18
  EmbeddingGemma-2-IQ4_NL.gguf 177640064 0.177640 PASS_EMBEDDING_CPU
19
  EmbeddingGemma-2-IQ3_M.gguf 145748608 0.145749 PASS_EMBEDDING_CPU
 
 
17
  EmbeddingGemma-2-IQ4_XS.gguf 169382528 0.169383 PASS_EMBEDDING_CPU
18
  EmbeddingGemma-2-IQ4_NL.gguf 177640064 0.177640 PASS_EMBEDDING_CPU
19
  EmbeddingGemma-2-IQ3_M.gguf 145748608 0.145749 PASS_EMBEDDING_CPU
20
+ EmbeddingGemma-2-IQ3_S.gguf 142545536 0.142546 PASS_EMBEDDING_CPU
reproducibility/manifest.md CHANGED
@@ -168,3 +168,11 @@ The fixed task evaluation uses the full MTEB STSBenchmark `test` split and the u
168
  - Quantization command: `llama-quantize --imatrix calibration/embeddinggemma-2/embeddinggemma-2-sts.imatrix.gguf embeddinggemma-2-text-BF16.gguf EmbeddingGemma-2-IQ3_M.gguf IQ3_M 8`.
169
  - CPU `llama-server --embedding --pooling mean -ngl 0` smoke passed with finite normalized 768-dimensional text vectors.
170
  - Build cgroup: `MemoryMax=32G`, `MemorySwapMax=0`; peak memory `883843072` bytes and swap peak `0` bytes.
 
 
 
 
 
 
 
 
 
168
  - Quantization command: `llama-quantize --imatrix calibration/embeddinggemma-2/embeddinggemma-2-sts.imatrix.gguf embeddinggemma-2-text-BF16.gguf EmbeddingGemma-2-IQ3_M.gguf IQ3_M 8`.
169
  - CPU `llama-server --embedding --pooling mean -ngl 0` smoke passed with finite normalized 768-dimensional text vectors.
170
  - Build cgroup: `MemoryMax=32G`, `MemorySwapMax=0`; peak memory `883843072` bytes and swap peak `0` bytes.
171
+ - IQ3_M public Hub commit: `893ef534c4a2828b7c6181260706d200f58a8ad6`; remote LFS SHA256 and size match the local file.
172
+
173
+ ## Incremental experiment: IQ3_S
174
+
175
+ - Artifact: `EmbeddingGemma-2-IQ3_S.gguf`; SHA256 `ce15d1c47c0b9bd3ab51921b8fbf4758b767fe8ae035a81b44783e37d01a652d`; decimal size `0.142546 GB`.
176
+ - Quantization command: `llama-quantize --imatrix calibration/embeddinggemma-2/embeddinggemma-2-sts.imatrix.gguf embeddinggemma-2-text-BF16.gguf EmbeddingGemma-2-IQ3_S.gguf IQ3_S 8`.
177
+ - CPU `llama-server --embedding --pooling mean -ngl 0` smoke passed with finite normalized 768-dimensional text vectors.
178
+ - Build cgroup: `MemoryMax=32G`, `MemorySwapMax=0`; peak memory `879149056` bytes and swap peak `0` bytes.