eaddario/Hammer2.1-7b-GGUF
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: eaddario/Hammer2.1-7b-GGUF
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-08-25 11:51:19 +08:00
commit 8b81781d2d
110 changed files with 15439 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
.gitattributes vendored Normal file
View File

@@ -0,0 +1,38 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.npy filter=lfs diff=lfs merge=lfs -text
*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pickle filter=lfs diff=lfs merge=lfs -text
*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text
*.gguf filter=lfs diff=lfs merge=lfs -text
*.logits filter=lfs diff=lfs merge=lfs -text
*.dat filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text

584
.gitignore vendored Normal file
View File

@@ -0,0 +1,584 @@
# https://github.com/github/gitignore
# -------------
# General Linux
# -------------
.directory
.fuse_hidden*
.nfs*
.Trash-*
*~
# -----------
# General OSX
# -----------
._*
.apdisk
.AppleDB
.AppleDesktop
.AppleDouble
.com.apple.timemachine.donotpresent
.DocumentRevisions-V100
.DS_Store
.fseventsd
.LSOverride
.Spotlight-V100
.TemporaryItems
.Trashes
.VolumeIcon.icns
Network Trash Folder
Temporary Items
# ---------------
# General Windows
# ---------------
[Dd]esktop.ini
*.cab
*.lnk
*.msi
*.msix
*.msm
*.msp
*.stackdump
$RECYCLE.BIN/
ehthumbs_vista.db
ehthumbs.db
Thumbs.db
Thumbs.db:encryptable
# -----------------
# General JetBrains
# -----------------
.idea_modules/
.idea/
*.iml
*.ipr
*.iws
# ---------------
# General VS Code
# ---------------
!.vscode/*.code-snippets
!.vscode/extensions.json
!.vscode/launch.json
!.vscode/settings.json
!.vscode/tasks.json
.history/
.vscode/
*.vsix
# ---------------------
# General Visual Studio
# ---------------------
__pycache__/
_NCrunch_*
_pkginfo.txt
_Pvt_Extensions
_ReSharper*/
_TeamCity*
_UpgradeReport_Files/
!?*.[Cc]ache/
!.vscode/extensions.json
!**/[Pp]ackages/build/
.*crunch*.local.xml
.builds
.cr/personal
.fake/
.ionide/
.localhistory/
.mfractor/
.ntvs_analysis.dat
.paket/paket.exe
.sass-cache/
.vs/
.vscode/*
.vshistory/
[Aa][Rr][Mm]/
[Aa][Rr][Mm]64/
[Bb]in/
[Bb]uild[Ll]og.*
[Dd]ebug/
[Dd]ebugPS/
[Dd]ebugPublic/
[Ee]xpress/
[Ll]og/
[Ll]ogs/
[Oo]bj/
[Rr]elease/
[Rr]eleasePS/
[Rr]eleases/
[Tt]est[Rr]esult*/
[Ww][Ii][Nn]32/
*_h.h
*_i.c
*_p.c
*_wpftmp.csproj
*- [Bb]ackup ([0-9]).rdl
*- [Bb]ackup ([0-9][0-9]).rdl
*- [Bb]ackup.rdl
*.[Cc]ache
*.[Pp]ublish.xml
*.[Rr]e[Ss]harper
*.appx
*.appxbundle
*.appxupload
*.aps
*.azurePubxml
*.bim_*.settings
*.bim.layout
*.binlog
*.btm.cs
*.btp.cs
*.build.csdef
*.cachefile
*.code-workspace
*.coverage
*.coveragexml
*.dbmdl
*.dbproj.schemaview
*.dotCover
*.DotSettings.user
*.dsp
*.dsw
*.e2e
*.GhostDoc.xml
*.gpState
*.ilk
*.iobj
*.ipdb
*.jfm
*.jmconfig
*.ldf
*.mdf
*.meta
*.mm.*
*.ncb
*.ndf
*.nuget.props
*.nuget.targets
*.nupkg
*.nvuser
*.obj
*.odx.cs
*.opendb
*.opensdf
*.opt
*.pch
*.pdb
*.pfx
*.pgc
*.pgd
*.pidb
*.plg
*.psess
*.publishproj
*.publishsettings
*.pubxml
*.pyc
*.rdl.data
*.rptproj.bak
*.rptproj.rsuser
*.rsp
*.rsuser
*.sap
*.sbr
*.scc
*.sdf
*.sln.docstates
*.sln.iml
*.snupkg
*.suo
*.svclog
*.tlb
*.tlh
*.tli
*.tlog
*.tmp
*.tmp_proj
*.tss
*.user
*.userosscache
*.userprefs
*.vbp
*.vbw
*.VC.db
*.VC.VC.opendb
*.VisualState.xml
*.vsp
*.vspscc
*.vspx
*.vssscc
*.xsd.cs
**/[Pp]ackages/*
**/*.DesktopClient/GeneratedArtifacts
**/*.DesktopClient/ModelManifest.xml
**/*.HTMLClient/GeneratedArtifacts
**/*.Server/GeneratedArtifacts
**/*.Server/ModelManifest.xml
~$*
$tf/
AppPackages/
artifacts/
ASALocalRun/
AutoTest.Net/
Backup*/
BenchmarkDotNet.Artifacts/
bld/
BundleArtifacts/
ClientBin/
coverage*.info
coverage*.json
coverage*.xml
csx/
dlldata.c
DocProject/buildhelp/
DocProject/Help/*.hhc
DocProject/Help/*.hhk
DocProject/Help/*.hhp
DocProject/Help/*.HxC
DocProject/Help/*.HxT
DocProject/Help/html
DocProject/Help/Html2
ecf/
FakesAssemblies/
FodyWeavers.xsd
Generated_Code/
healthchecksdb
ipch/
MigrationBackup/
mono_crash.*
nCrunchTemp_*
node_modules/
nunit-*.xml
OpenCover/
orleans.codegen.cs
Package.StoreAssociation.xml
paket-files/
project.fragment.lock.json
project.lock.json
publish/
PublishScripts/
rcf/
ScaffoldingReadMe.txt
ServiceFabricBackup/
StyleCopReport.xml
TestResult.xml
UpgradeLog*.htm
UpgradeLog*.XML
x64/
x86/
# ----------------------
# General Archived Files
# ----------------------
*.7z
*.bz2
*.bzip
*.bzip2
*.deb
*.dmg
*.egg
*.gem
*.gz
*.gzip
*.iso
*.jar
*.lzma
*.rar
*.rpm
*.tar
*.tgz
*.txz
*.tzst
*.xar
*.xpi
*.xz
*.zip
*.zst
# -----
# C/C++
# -----
.tmp_versions/
*.a
*.app
*.cmd
*.d
*.dll
*.dSYM/
*.dylib
*.elf
*.exe
*.exp
*.gch
*.hex
*.i*86
*.idb
*.ko
*.la
*.lai
*.lib
*.lo
*.map
*.mod*
*.o
*.out
*.slo
*.so
*.so.*
*.su
*.x86_64
dkms.conf
Mkfile.old
Module.symvers
modules.order
# ----
# CUDA
# ----
*.cubin
*.fatbin
*.gpu
*.i
*.ii
*.ptx
# --
# Go
# --
.env
*.exe~
*.test
go.work
go.work.sum
# ----
# Java
# ----
*.class
*.ctxt
*.ear
*.hprof
*.nar
*.tar.gz
*.war
hs_err_pid*
replay_pid*
# -----
# Julia
# -----
*.jl.*.cov
*.jl.cov
*.jl.mem
deps/build.log
deps/deps.jl
deps/downloads/
deps/src/
deps/usr/
docs/build/
docs/site/
Manifest.toml
# ------------------------------
# JavaScript / Node / TypeScript
# ------------------------------
.cache
.cache/
.docusaurus
.dynamodb/
.env.development.local
.env.local
.env.production.local
.env.test.local
.eslintcache
.fusebox/
.grunt
.lock-wscript
.next
.node_repl_history
.npm
.nuxt
.nyc_output
.parcel-cache
.pnp.*
.pnpm-debug.log*
.rpt2_cache/
.rts2_cache_cjs/
.rts2_cache_es/
.rts2_cache_umd/
.serverless/
.stylelintcache
.temp
.tern-port
.vscode-test
.vuepress/dist
.yarn-integrity
.yarn/build-state.yml
.yarn/cache
.yarn/install-state.gz
.yarn/unplugged
*.lcov
*.pid
*.pid.lock
*.seed
*.tsbuildinfo
bower_components
dist
jspm_packages/
lerna-debug.log*
lib-cov
logs
npm-debug.log*
out
pids
report.[0-9]*.[0-9]*.[0-9]*.[0-9]*.json
web_modules/
yarn-debug.log*
yarn-error.log*
# ------
# Python
# ------
__pypackages__/
.coverage
.coverage.*
.dmypy.json
.eggs/
.hypothesis/
.installed.cfg
.ipynb_checkpoints
.mypy_cache/
.nox/
.pdm-build/
.pdm-python
.pdm.toml
.pybuilder/
.pyre/
.pytest_cache/
.Python
.python-version
.pytype/
.ropeproject
.scrapy
.spyderproject
.spyproject
.tox/
.venv
.webassets-cache
*.cover
*.egg-info/
*.manifest
*.mo
*.pot
*.py,cover
*.py[cod]
*.sage.py
*.spec
*$py.class
/site
build/
celerybeat-schedule
celerybeat.pid
cover/
coverage.xml
cython_debug/
db.sqlite3
db.sqlite3-journal
develop-eggs/
dist/
dmypy.json
docs/_build/
downloads/
eggs/
env.bak/
env/
htmlcov/
instance/
ipython_config.py
lib/
lib64/
local_settings.py
MANIFEST
nosetests.xml
parts/
pip-delete-this-directory.txt
pip-log.txt
profile_default/
sdist/
share/python-wheels/
target/
var/
venv.bak/
venv/
wheels/
# ----
# Rust
# ----
**/*.rs.bk
Cargo.lock
debug/
# -----
# Scala
# -----
# -----
# CMake
# -----
_deps
cmake_install.cmake
CMakeCache.txt
CMakeFiles
CMakeLists.txt.user
CMakeScripts
CMakeUserPresets.json
compile_commands.json
CTestTestfile.cmake
install_manifest.txt
Makefile
Testing
# ------
# Gradle
# ------
!gradle-wrapper.jar
!gradle-wrapper.properties
!src/**/build/
.classpath
.gradle
.gradletasknamecache
.project
**/build/
gradle-app.setting
# -----
# Maven
# -----
.mvn/
buildNumber.properties
dependency-reduced-pom.xml
pom.xml.next
pom.xml.releaseBackup
pom.xml.tag
pom.xml.versionsBackup
release.properties
# ---------
# Terraform
# ---------
.terraform.tfstate.lock.info
.terraform/
.terraformrc
*_override.tf
*_override.tf.json
*.tfstate
*.tfstate.*
*.tfvars
*.tfvars.json
crash.*.log
crash.log
override.tf
override.tf.json
terraform.rc

3
Hammer2.1-7b-F16.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:d414b3f9607335467c199f2378d25d8f19bcb465376412aca4c2742f5a90452d
size 15232123808

3
Hammer2.1-7b-IQ3_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:f463f9b3bec8377f80cea123711fcdab55613f4f0ff352091e9280ba048dbe4b
size 3483159680

3
Hammer2.1-7b-IQ3_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:97cb3efe4747109275b82568ad6b4751b035befa08dc22d1e8e80a5dc3ea6476
size 3254777856

3
Hammer2.1-7b-IQ4_NL.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:88a7050e28853a0249d05c75f37747628867822c182d302da4c101eddb3ea95e
size 4174122624

3
Hammer2.1-7b-Q3_K_L.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:033d0484dc19b37492f9b96022d9777702b5809a749127c2a7bd1e4f8c5cb536
size 3535142016

3
Hammer2.1-7b-Q3_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:676d89f69276fe589b30c458f509aac0f50e81590ff4fbcfa4c729d30bdc147f
size 3367955584

3
Hammer2.1-7b-Q3_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:f73cf6d45841ed732d1b000b878ba00c1043f671c76a959a66d3002f038d458c
size 3135327168

3
Hammer2.1-7b-Q4_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:cd698cb286d02e407ef41b2304f4b7fd1f7d45eece1780120016e5fbc41e8020
size 4178136704

3
Hammer2.1-7b-Q4_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:d01de17d7bf4309e0e5aad479c6b2d69e56ba9ca9814c16d1318dd814308090c
size 4056108672

3
Hammer2.1-7b-Q5_K_M.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:eb13f56264f87a5e1bf15d82cb7bc32218b83995fe562da4cce6f659bc18684f
size 5093254208

3
Hammer2.1-7b-Q5_K_S.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:d7940c008124ba606a6c60381c78f24c92937e3a547d61e1aa7332295cba0ea7
size 4963599424

3
Hammer2.1-7b-Q6_K.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:a566f0f780e41ae072ba7f40cf2d660c12ad0fc8fcd54de0ea9d50c90d913f72
size 6226368096

3
Hammer2.1-7b-Q8_0.gguf Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:ac8f52cffb377844263f785b4a18409cb7e7dcac1882952d802df2b5af851854
size 7265396096

143
README.md Normal file
View File

@@ -0,0 +1,143 @@
---
base_model:
- MadeAgents/Hammer2.1-7b
datasets:
- eaddario/imatrix-calibration
language:
- en
license:
- cc-by-nc-4.0
pipeline_tag: text-generation
tags:
- gguf
- quant
- experimental
---
# Experimental layer-wise quantization of MadeAgents/Hammer2.1-7b
Using [LLaMA C++](<https://github.com/ggerganov/llama.cpp>) release [b5170](<https://github.com/ggerganov/llama.cpp/releases/tag/b5170>) for quantization.
Original model: [MadeAgents/Hammer2.1-7b](https://huggingface.co/MadeAgents/Hammer2.1-7b)
From the original model creators:
> Hammer refers to a series of lightweight Large Action Models. Currently, we are releasing Hammer 2.1 models ([0.5B](https://huggingface.co/MadeAgents/Hammer2.1-0.5b), [1.5B](https://huggingface.co/MadeAgents/Hammer2.1-1.5b), [3B](https://huggingface.co/MadeAgents/Hammer2.1-3b), and [7B](https://huggingface.co/MadeAgents/Hammer2.1-7b)) with strong function calling capability. These models are based on the Qwen 2.5 coder series and utilize [function masking techniques](https://arxiv.org/abs/2410.04587) and other advanced technologies. Hammer 2.1 series bring significant enhancements, while still maintaining the basic functionality of Hammer 2.0's Single-Turn interaction and further strengthening other capabilities.
>
> The Hammer 2.1 models, fine-tuned from the Qwen 2.5 coder series, inherit Hammer 2.0's advantages and are enhanced as follows:
> - Multi-Step Function Calling: The assistant can perform multiple internal function calls to handle a single user request, actively planning and gathering information to fulfill complex tasks.
> - Multi-Turn Function Calling: Enables continuous and context-aware interactions over multiple exchanges, with each turn potentially containing multiple steps, for a more natural conversation experience.
> - Enhanced Irrelevant Information Inspection: Better at identifying when provided functions are irrelevant to a user query, by providing a non-function call response.
# PLEASE READ THIS BEFORE USING THESE EXPERIMENTAL VERSIONS!
An area of personal interest is finding ways to optimize the inference performance of LLMs when deployed in resource-constrained environments like commodity hardware, desktops, laptops, mobiles, edge devices, etc. There are many approaches to accomplish this, including architecture simplification and knowledge distillation, but my focus has been primarily on quantization and pruning.
The method used to produce these experimental versions is covered in [Squeezing Tensor Bits: the quest for smaller LLMs](<https://medium.com/@eaddario/squeezing-tensor-bits-the-quest-for-smaller-llms-86b23bd052ca>), but at a high level it involves using custom versions of `llama-imatrix` and `llama-quantize` to identify the influential tensors, and quantize the most important layers to higher bit precision and the less important to lower bits. This process was partly inspired by Dumitru's et al [Layer-Wise Quantization: A Pragmatic and Effective Method for Quantizing LLMs Beyond Integer Bit-Levels](<https://arxiv.org/abs/2406.17415>).
There’re two pull requests ([imatrix](<https://github.com/ggml-org/llama.cpp/pull/12718>) & [quantize](<https://github.com/ggml-org/llama.cpp/pull/12511>)) to merge these changes back into the core llama.cpp project. This may or may not ever happen so, until then, the modified versions will be available on [GitHub](<https://github.com/EAddario/llama.cpp>).
For testing and comparison I'd normally use models produced by [Unsloth](<https://huggingface.co/unsloth>) ([Daniel and Michael Han](<https://unsloth.ai/>) do some really advanced level stuff!) and [Bartowski](<https://huggingface.co/bartowski>) (see credits below), but they don't provide GGUF versions of this model, so all tests and comparisons are done against naive quantizations obtained by simply running `llama-quantize` with no further optimization.
All experimental versions were generated using an appropriate imatrix created from calibration datasets available at [eaddario/imatrix-calibration](<https://huggingface.co/datasets/eaddario/imatrix-calibration>). At its core, an Importance Matrix (imatrix) is a table or, more broadly, a structured representation that scores the relative importance of different features or parameters in a machine learning model. It essentially quantifies the "impact" each feature has on a specific outcome, prediction, or relationship being modeled, and it helps to counterbalance the negative effects of quantization and pruning.
The process to generate these models is roughly as follows:
1. Convert the the original model's tensors to [GGUF](<https://huggingface.co/docs/hub/en/gguf>) F16*
2. Estimate the Perplexity score for the F16 model (baseline) using the [wikitext-2-raw-v1](<https://huggingface.co/datasets/Salesforce/wikitext/tree/main/wikitext-2-raw-v1>) dataset, and save the [logits](<https://huggingface.co/eaddario/Hammer2.1-7b-GGUF/tree/main/logits>)
3. Generate an [imatrix](<https://huggingface.co/eaddario/Hammer2.1-7b-GGUF/tree/main/imatrix>) from selected calibration datasets
4. Determine tensor and layer Importance Score contribution using a modified version of `llama-imatrix`
5. Select an appropiate quant level for each tensor using a modified version of `llama-quantize`
6. Calculate Perplexity, KL Divergence, ARC (Easy+Challenge), HellaSwag, MMLU, Truthful QA and WinoGrande scores for each quantized model
7. Keep versions with the best scores
8. Repeat until all desired quants are created. I find that quantizations below Q3/IQ3 are not fit for my purposes and therefore do not usually generate them, but happy to provide other quants on request.
*[BF16](<https://en.wikipedia.org/wiki/Bfloat16_floating-point_format>) would be preferred, but Apple's GPUs don't support it yet, and therefore any operations are executed in the CPU, making it unacceptably slow. This is expected to change in the near term but until then, if you are using Apple kit avoid using any models tagged BF16
# Models
### Sizes (in GB)
| Model | Naive | Repo | Shrinkage |
| ------------------------------------------------- | ----: | ---: | --------: |
| [Hammer2.1-7b-IQ3_M](./Hammer2.1-7b-IQ3_M.gguf) | 3.57 | 3.48 | 2.5% |
| [Hammer2.1-7b-IQ3_S](./Hammer2.1-7b-IQ3_S.gguf) | 3.50 | 3.25 | 7.1% |
| [Hammer2.1-7b-IQ4_NL](./Hammer2.1-7b-IQ4_NL.gguf) | 4.44 | 4.17 | 6.1% |
| [Hammer2.1-7b-Q3_K_L](./Hammer2.1-7b-Q3_K_L.gguf) | 4.09 | 3.54 | 13.4% |
| [Hammer2.1-7b-Q3_K_M](./Hammer2.1-7b-Q3_K_M.gguf) | 3.81 | 3.37 | 11.5% |
| [Hammer2.1-7b-Q3_K_S](./Hammer2.1-7b-Q3_K_S.gguf) | 3.49 | 3.14 | 10.0% |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 4.68 | 4.18 | 10.7% |
| [Hammer2.1-7b-Q4_K_S](./Hammer2.1-7b-Q4_K_S.gguf) | 4.46 | 4.06 | 9.0% |
| [Hammer2.1-7b-Q5_K_M](./Hammer2.1-7b-Q5_K_M.gguf) | 5.44 | 5.09 | 6.4% |
| [Hammer2.1-7b-Q5_K_S](./Hammer2.1-7b-Q5_K_S.gguf) | 5.31 | 4.96 | 6.6% |
| [Hammer2.1-7b-Q6_K](./Hammer2.1-7b-Q6_K.gguf) | 6.25 | 6.23 | 0.3% |
| [Hammer2.1-7b-Q8_0](./Hammer2.1-7b-Q8_0.gguf) | 8.10 | 7.27 | 10.2% |
### Perplexity and KL Divergence scores
| Model | μPPL | 𝜌PPL | μKLD | RMS Δp |
| ------------------------------------------------- | ------------------: | -----: | -----------------: | -----------: |
| [Hammer2.1-7b-IQ3_M](./Hammer2.1-7b-IQ3_M.gguf) | 9.827809 ±0.071236 | 98.97% | 0.059970 ±0.000226 | 6.574 ±0.032 |
| [Hammer2.1-7b-IQ3_S](./Hammer2.1-7b-IQ3_S.gguf) | 9.942560 ±0.071243 | 98.64% | 0.079935 ±0.000289 | 7.445 ±0.038 |
| [Hammer2.1-7b-IQ4_NL](./Hammer2.1-7b-IQ4_NL.gguf) | 9.585928 ±0.068060 | 99.51% | 0.026627 ±0.000108 | 4.346 ±0.026 |
| [Hammer2.1-7b-Q3_K_L](./Hammer2.1-7b-Q3_K_L.gguf) | 10.010842 ±0.071669 | 98.55% | 0.080230 ±0.000317 | 7.556 ±0.040 |
| [Hammer2.1-7b-Q3_K_M](./Hammer2.1-7b-Q3_K_M.gguf) | 10.109660 ±0.072549 | 98.44% | 0.086444 ±0.000342 | 7.845 ±0.041 |
| [Hammer2.1-7b-Q3_K_S](./Hammer2.1-7b-Q3_K_S.gguf) | 10.224307 ±0.073520 | 98.16% | 0.101951 ±0.000397 | 8.487 ±0.044 |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 9.513981 ±0.067589 | 99.56% | 0.023792 ±0.000107 | 4.057 ±0.026 |
| Hammer2.1-7b-Q4_K_M (naive) | 9.469517 ±0.067164 | 99.70% | 0.016770 ±0.000081 | 3.364 ±0.022 |
| [Hammer2.1-7b-Q4_K_S](./Hammer2.1-7b-Q4_K_S.gguf) | 9.529467 ±0.067693 | 99.53% | 0.025429 ±0.000116 | 4.191 ±0.026 |
| [Hammer2.1-7b-Q5_K_M](./Hammer2.1-7b-Q5_K_M.gguf) | 9.411472 ±0.066924 | 99.88% | 0.005881 ±0.000035 | 2.028 ±0.015 |
| [Hammer2.1-7b-Q5_K_S](./Hammer2.1-7b-Q5_K_S.gguf) | 9.410802 ±0.066883 | 99.88% | 0.006252 ±0.000036 | 2.085 ±0.015 |
| [Hammer2.1-7b-Q6_K](./Hammer2.1-7b-Q6_K.gguf) | 9.384561 ±0.066615 | 99.96% | 0.001779 ±0.000018 | 1.101 ±0.012 |
| [Hammer2.1-7b-Q8_0](./Hammer2.1-7b-Q8_0.gguf) | 9.379380 ±0.066571 | 99.98% | 0.000692 ±0.000012 | 0.706 ±0.010 |
| [Hammer2.1-7b-F16](./Hammer2.1-7b-F16.gguf) | 9.366577 ±0.066397 | 100% | N/A | N/A |
### ARC, HellaSwag, MMLU, Truthful QA and WinoGrande scores
Scores generated using [llama-perplexity](<https://github.com/ggml-org/llama.cpp/tree/master/examples/perplexity>) with 750 tasks per test, and a context size of 768 tokens.
For the test data used in the generation of these scores, follow the appropiate links: [HellaSwag](<https://github.com/klosax/hellaswag_text_data>), [ARC, MMLU, Truthful QA](<https://huggingface.co/datasets/ikawrakow/validation-datasets-for-llama.cpp/tree/main>) and [WinoGrande](<https://huggingface.co/datasets/ikawrakow/winogrande-eval-for-llama.cpp/tree/main>)
| Model | ARC | HellaSwag | MMLU | Truthful QA | WinoGrande | Avg Score |
| ------------------------------------------------- | --------------: | --------: | --------------: | --------------: | --------------: | --------: |
| [Hammer2.1-7b-IQ3_M](./Hammer2.1-7b-IQ3_M.gguf) | 56.5333 ±1.8113 | 71.46 | 38.0000 ±1.7736 | 29.8667 ±1.6723 | 70.6667 ±1.6636 | 53.31 |
| [Hammer2.1-7b-IQ3_S](./Hammer2.1-7b-IQ3_S.gguf) | 55.8667 ±1.8143 | 70.26 | 38.6667 ±1.7794 | 30.9333 ±1.6889 | 70.4000 ±1.6680 | 53.23 |
| [Hammer2.1-7b-IQ4_NL](./Hammer2.1-7b-IQ4_NL.gguf) | 57.3333 ±1.8072 | 71.20 | 38.5333 ±1.7783 | 30.1333 ±1.6766 | 69.8667 ±1.6766 | 53.41 |
| [Hammer2.1-7b-Q3_K_L](./Hammer2.1-7b-Q3_K_L.gguf) | 57.2000 ±1.8079 | 69.86 | 36.6667 ±1.7608 | 29.6000 ±1.6680 | 70.9333 ±1.6591 | 52.85 |
| [Hammer2.1-7b-Q3_K_M](./Hammer2.1-7b-Q3_K_M.gguf) | 55.7333 ±1.8149 | 69.87 | 37.2000 ±1.7661 | 29.8667 ±1.6723 | 70.4000 ±1.6680 | 52.61 |
| [Hammer2.1-7b-Q3_K_S](./Hammer2.1-7b-Q3_K_S.gguf) | 56.5333 ±1.8113 | 68.80 | 37.3333 ±1.7674 | 29.0667 ±1.6591 | 69.3333 ±1.6849 | 52.21 |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 57.8667 ±1.8042 | 71.20 | 38.0000 ±1.7736 | 29.8667 ±1.6723 | 69.8667 ±1.6766 | 53.36 |
| Hammer2.1-7b-Q4_K_M (naive) | 57.8313 ±1.8080 | 73.47 | 35.0667 ±1.7436 | 33.9683 ±2.6727 | 70.8000 ±1.6614 | 54.23 |
| [Hammer2.1-7b-Q4_K_S](./Hammer2.1-7b-Q4_K_S.gguf) | 57.8667 ±1.8042 | 71.07 | 37.8667 ±1.7724 | 29.6000 ±1.6680 | 69.6000 ±1.6807 | 53.20 |
| [Hammer2.1-7b-Q5_K_M](./Hammer2.1-7b-Q5_K_M.gguf) | 58.0000 ±1.8034 | 71.33 | 38.6667 ±1.7794 | 30.2667 ±1.6787 | 70.6667 ±1.6636 | 53.79 |
| [Hammer2.1-7b-Q5_K_S](./Hammer2.1-7b-Q5_K_S.gguf) | 58.4000 ±1.8010 | 71.47 | 38.8000 ±1.7805 | 30.5333 ±1.6828 | 70.6667 ±1.6636 | 53.97 |
| [Hammer2.1-7b-Q6_K](./Hammer2.1-7b-Q6_K.gguf) | 58.8000 ±1.7984 | 72.00 | 38.9333 ±1.7816 | 29.7333 ±1.6702 | 70.4000 ±1.6680 | 53.97 |
| [Hammer2.1-7b-Q8_0](./Hammer2.1-7b-Q8_0.gguf) | 58.5333 ±1.8002 | 72.00 | 38.9333 ±1.7816 | 30.1333 ±1.6766 | 70.2667 ±1.6702 | 53.97 |
| [Hammer2.1-7b-F16](./Hammer2.1-7b-F16.gguf) | 58.9333 ±1.7976 | 72.13 | 39.0667 ±1.7827 | 30.4000 ±1.6807 | 70.5333 ±1.6658 | 54.21 |
### Tokens per Second - Benchmarks
Scores generated using [llama-bench](https://github.com/ggml-org/llama.cpp/tree/master/examples/llama-bench). Naive Q4_K_M quantization included for comparison.
| model | size | params | backend | threads | test | t/s |
| ------------------------------------------------- | -------: | -----: | ---------- | ------: | ------------: | ------------: |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 3.89 GiB | 7.61 B | Metal,BLAS | 6 | pp512 | 336.11 ± 0.60 |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 3.89 GiB | 7.61 B | Metal,BLAS | 6 | tg128 | 29.32 ± 0.14 |
| [Hammer2.1-7b-Q4_K_M](./Hammer2.1-7b-Q4_K_M.gguf) | 3.89 GiB | 7.61 B | Metal,BLAS | 6 | pp1024+tg1024 | 48.00 ± 0.16 |
| Hammer2.1-7b-Q4_K_M (naive) | 4.35 GiB | 7.61 B | Metal,BLAS | 6 | pp512 | 355.08 ± 0.19 |
| Hammer2.1-7b-Q4_K_M (naive) | 4.35 GiB | 7.61 B | Metal,BLAS | 6 | tg128 | 28.21 ± 0.00 |
| Hammer2.1-7b-Q4_K_M (naive) | 4.35 GiB | 7.61 B | Metal,BLAS | 6 | pp1024+tg1024 | 45.92 ± 1.18 |
# Metrics used
**[Perplexity](<https://huggingface.co/docs/transformers/en/perplexity>):** one of the key metrics used in NLP evaluation. It measures the quality of a language model by evaluating how well it predicts the next token given a particular sequence of words. A PPL of **1** indicates an exact match between predicted and actual, whereas values greater than one indicate a degree of "surprise" the generated token differs from the expected.
**[Kullback–Leibler (KL) Divergence](<https://en.wikipedia.org/wiki/Kullback–Leibler_divergence>):** a statistical measure of how much a probability distribution differs from another. When quantizing models (or altering the original tensors in any way for that matter), the closest we can preserve the weights' probability distribution to the original model the better, thus the closest to **0** the better.
**[AI2 Reasoning Challenge (ARC)](<https://leaderboard.allenai.org/arc/submissions/get-started>):** a benchmark to evaluate the ability of AI models to answer complex science questions that require logical reasoning beyond pattern matching.
**[HellaSwag](<https://rowanzellers.com/hellaswag/>):** the Harder Endings, Longer contexts, and Low-shot Activities for Situations With Adversarial Generations (bit of a mouthful!) is a benchmark designed to test commonsense natural language inference. It requires the model to predict the most likely ending of a sentence.
**[MMLU](<https://github.com/hendrycks/test>):** the Massive Multitask Language Understanding evaluates LLMs’ general knowledge and problem-solving abilities across 57 subjects, including elementary mathematics, US history, computer science, and law.
**[Truthful QA](<https://github.com/sylinrl/TruthfulQA>):** evaluates how well LLMs generate truthful responses to questions. It identifies whether AI models can avoid generating false or misleading information, particularly in areas where human knowledge is prone to misconceptions.
**[Winogrande](<https://winogrande.allenai.org/>):** based on the [Winograd Schema Challenge](<https://cdn.aaai.org/ocs/4492/4492-21843-1-PB.pdf>), is a natural language understanding task requiring models to resolve ambiguities in sentences involving pronoun references.
## Credits
A big **Thank You!** to [Colin Kealty](<https://huggingface.co/bartowski>) for the many contributions and for being one of the best sources of high quality quantized models available in Hugginface, and a really big ***Thank You!*** to [Georgi Gerganov](<https://github.com/ggerganov>) for his amazing work with **llama.cpp** and the **ggml/gguf** libraries.

3
imatrix/imatrix-Hammer2.1-7b-micro.dat Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:7d33304f929548e097be5afdc04c00ab91a729ec3cfbb30639fae5555f56dd65
size 4551051

3
imatrix/imatrix-Hammer2.1-7b-small.dat Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:7866bac797b7678785e273994874f59fbf2a331dfc0e0e5fd0ec7fcecea754ff
size 4551051

3
imatrix/imatrix-Hammer2.1-7b-tiny.dat Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:8e2b2c0fa25b7286cea883300aa904f099ffd60e8630da388bc492c8ed6be0a1
size 4551050

3
logits/Hammer2.1-7b-F16.logits Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:5c2428621962d4b5693e38f2e0a470cce82b7b65283f9e0b753d5aaabfe4e6aa
size 45194911608

13
scores/Hammer2.1-7b-F16.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 29 key-value pairs and 339 tensors from ./Hammer2.1-7b-F16.gguf (version GGUF V3 (latest))
Final result: 58.9333 +/- 1.7976
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 6229.12 ms
llama_perf_context_print: prompt eval time = 102811.05 ms / 35972 tokens ( 2.86 ms per token, 349.88 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 103951.36 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-F16.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 29 key-value pairs and 339 tensors from ./Hammer2.1-7b-F16.gguf (version GGUF V3 (latest))
750 72.13333333% [68.8181%, 75.2230%]
llama_perf_context_print: load time = 488.78 ms
llama_perf_context_print: prompt eval time = 371768.51 ms / 126038 tokens ( 2.95 ms per token, 339.02 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 377711.49 ms / 126039 tokens
ggml_metal_free: deallocating

1025
scores/Hammer2.1-7b-F16.md Normal file
View File

File diff suppressed because it is too large Load Diff

13
scores/Hammer2.1-7b-F16.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 29 key-value pairs and 339 tensors from ./Hammer2.1-7b-F16.gguf (version GGUF V3 (latest))
Final result: 39.0667 +/- 1.7827
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 527.95 ms
llama_perf_context_print: prompt eval time = 194228.90 ms / 67719 tokens ( 2.87 ms per token, 348.66 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 196042.71 ms / 67720 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-F16.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 29 key-value pairs and 339 tensors from ./Hammer2.1-7b-F16.gguf (version GGUF V3 (latest))
Final result: 30.4000 +/- 1.6807
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 508.17 ms
llama_perf_context_print: prompt eval time = 146171.42 ms / 49696 tokens ( 2.94 ms per token, 339.98 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 148488.63 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-F16.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 29 key-value pairs and 339 tensors from ./Hammer2.1-7b-F16.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.5333 +/- 1.6658
llama_perf_context_print: load time = 536.90 ms
llama_perf_context_print: prompt eval time = 61935.07 ms / 21448 tokens ( 2.89 ms per token, 346.30 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 62681.37 ms / 21449 tokens
ggml_metal_free: deallocating

1029
scores/Hammer2.1-7b-IQ3_M.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-IQ3_S.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-IQ4_NL.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q3_K_L.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q3_K_M.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q3_K_S.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q4_K_M.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q4_K_S.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q5_K_M.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q5_K_S.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q6_K.md Normal file
View File

File diff suppressed because it is too large Load Diff

1029
scores/Hammer2.1-7b-Q8_0.md Normal file
View File

File diff suppressed because it is too large Load Diff

13
scores/Hammer2.1-7b-iq3_m.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_M.gguf (version GGUF V3 (latest))
Final result: 56.5333 +/- 1.8113
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1682.53 ms
llama_perf_context_print: prompt eval time = 106353.34 ms / 35972 tokens ( 2.96 ms per token, 338.23 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 107373.11 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-iq3_m.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_M.gguf (version GGUF V3 (latest))
750 71.46666667% [68.1319%, 74.5827%]
llama_perf_context_print: load time = 253.78 ms
llama_perf_context_print: prompt eval time = 375081.86 ms / 126038 tokens ( 2.98 ms per token, 336.03 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 380601.65 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-iq3_m.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_M.gguf (version GGUF V3 (latest))
Final result: 38.0000 +/- 1.7736
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 252.71 ms
llama_perf_context_print: prompt eval time = 195840.26 ms / 67719 tokens ( 2.89 ms per token, 345.79 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 197498.82 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-iq3_m.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.827809 ± 0.071236
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 98.97%
Mean ln(PPL(Q)/PPL(base)) : 0.048068 ± 0.001042
Mean PPL(Q)/PPL(base) : 1.049242 ± 0.001093
Mean PPL(Q)-PPL(base) : 0.461232 ± 0.011000
====== KL divergence statistics ======
Mean KLD: 0.059970 ± 0.000226
Maximum KLD: 5.830572
99.9% KLD: 0.899464
99.0% KLD: 0.388867
99.0% KLD: 0.388867
Median KLD: 0.039933
10.0% KLD: 0.000901
5.0% KLD: 0.000182
1.0% KLD: 0.000013
Minimum KLD: -0.000098
====== Token probability statistics ======
Mean Δp: -0.217 ± 0.017 %
Maximum Δp: 90.326%
99.9% Δp: 34.151%
99.0% Δp: 18.912%
95.0% Δp: 9.734%
90.0% Δp: 5.703%
75.0% Δp: 1.068%
Median Δp: -0.002%
25.0% Δp: -1.292%
10.0% Δp: -6.110%
5.0% Δp: -10.513%
1.0% Δp: -22.188%
0.1% Δp: -43.583%
Minimum Δp: -81.978%
RMS Δp : 6.574 ± 0.032 %
Same top p: 87.097 ± 0.087 %

13
scores/Hammer2.1-7b-iq3_m.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_M.gguf (version GGUF V3 (latest))
Final result: 29.8667 +/- 1.6723
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 252.53 ms
llama_perf_context_print: prompt eval time = 151286.74 ms / 49696 tokens ( 3.04 ms per token, 328.49 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 153386.94 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-iq3_m.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_M.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.6667 +/- 1.6636
llama_perf_context_print: load time = 254.13 ms
llama_perf_context_print: prompt eval time = 63679.99 ms / 21448 tokens ( 2.97 ms per token, 336.81 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 64320.92 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-iq3_s.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_S.gguf (version GGUF V3 (latest))
Final result: 55.8667 +/- 1.8143
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1546.80 ms
llama_perf_context_print: prompt eval time = 106501.37 ms / 35972 tokens ( 2.96 ms per token, 337.76 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 107504.12 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-iq3_s.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_S.gguf (version GGUF V3 (latest))
750 70.26666667% [66.8989%, 73.4279%]
llama_perf_context_print: load time = 251.76 ms
llama_perf_context_print: prompt eval time = 375564.87 ms / 126038 tokens ( 2.98 ms per token, 335.60 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 381034.82 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-iq3_s.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_S.gguf (version GGUF V3 (latest))
Final result: 38.6667 +/- 1.7794
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 252.58 ms
llama_perf_context_print: prompt eval time = 196145.96 ms / 67719 tokens ( 2.90 ms per token, 345.25 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 197800.20 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-iq3_s.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.942560 ± 0.071243
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 98.64%
Mean ln(PPL(Q)/PPL(base)) : 0.059677 ± 0.001178
Mean PPL(Q)/PPL(base) : 1.061493 ± 0.001251
Mean PPL(Q)-PPL(base) : 0.575982 ± 0.012339
====== KL divergence statistics ======
Mean KLD: 0.079935 ± 0.000289
Maximum KLD: 5.765717
99.9% KLD: 1.150883
99.0% KLD: 0.493128
99.0% KLD: 0.493128
Median KLD: 0.053589
10.0% KLD: 0.001480
5.0% KLD: 0.000323
1.0% KLD: 0.000027
Minimum KLD: -0.000038
====== Token probability statistics ======
Mean Δp: -0.885 ± 0.019 %
Maximum Δp: 91.523%
99.9% Δp: 34.667%
99.0% Δp: 19.093%
95.0% Δp: 9.310%
90.0% Δp: 5.250%
75.0% Δp: 0.738%
Median Δp: -0.033%
25.0% Δp: -2.046%
10.0% Δp: -7.921%
5.0% Δp: -13.116%
1.0% Δp: -26.866%
0.1% Δp: -52.295%
Minimum Δp: -90.266%
RMS Δp : 7.445 ± 0.038 %
Same top p: 85.475 ± 0.091 %

13
scores/Hammer2.1-7b-iq3_s.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_S.gguf (version GGUF V3 (latest))
Final result: 30.9333 +/- 1.6889
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 248.10 ms
llama_perf_context_print: prompt eval time = 151449.03 ms / 49696 tokens ( 3.05 ms per token, 328.14 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 153545.12 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-iq3_s.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ3_S.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.4000 +/- 1.6680
llama_perf_context_print: load time = 250.05 ms
llama_perf_context_print: prompt eval time = 63750.53 ms / 21448 tokens ( 2.97 ms per token, 336.44 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 64386.61 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-iq4_nl.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ4_NL.gguf (version GGUF V3 (latest))
Final result: 57.3333 +/- 1.8072
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1954.91 ms
llama_perf_context_print: prompt eval time = 110236.28 ms / 35972 tokens ( 3.06 ms per token, 326.32 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 111260.63 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-iq4_nl.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ4_NL.gguf (version GGUF V3 (latest))
750 71.20000000% [67.8576%, 74.3263%]
llama_perf_context_print: load time = 270.76 ms
llama_perf_context_print: prompt eval time = 388008.19 ms / 126038 tokens ( 3.08 ms per token, 324.83 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 393490.42 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-iq4_nl.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ4_NL.gguf (version GGUF V3 (latest))
Final result: 38.5333 +/- 1.7783
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 262.83 ms
llama_perf_context_print: prompt eval time = 202888.48 ms / 67719 tokens ( 3.00 ms per token, 333.77 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 204561.85 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-iq4_nl.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.585928 ± 0.068060
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 99.51%
Mean ln(PPL(Q)/PPL(base)) : 0.023148 ± 0.000703
Mean PPL(Q)/PPL(base) : 1.023418 ± 0.000719
Mean PPL(Q)-PPL(base) : 0.219350 ± 0.006862
====== KL divergence statistics ======
Mean KLD: 0.026627 ± 0.000108
Maximum KLD: 2.930657
99.9% KLD: 0.451824
99.0% KLD: 0.171826
99.0% KLD: 0.171826
Median KLD: 0.017210
10.0% KLD: 0.000451
5.0% KLD: 0.000095
1.0% KLD: 0.000007
Minimum KLD: -0.000060
====== Token probability statistics ======
Mean Δp: -0.523 ± 0.011 %
Maximum Δp: 54.223%
99.9% Δp: 22.680%
99.0% Δp: 11.163%
95.0% Δp: 5.267%
90.0% Δp: 2.899%
75.0% Δp: 0.419%
Median Δp: -0.014%
25.0% Δp: -1.188%
10.0% Δp: -4.623%
5.0% Δp: -7.618%
1.0% Δp: -15.106%
0.1% Δp: -30.059%
Minimum Δp: -89.467%
RMS Δp : 4.346 ± 0.026 %
Same top p: 91.424 ± 0.073 %

13
scores/Hammer2.1-7b-iq4_nl.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ4_NL.gguf (version GGUF V3 (latest))
Final result: 30.1333 +/- 1.6766
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 260.43 ms
llama_perf_context_print: prompt eval time = 156558.33 ms / 49696 tokens ( 3.15 ms per token, 317.43 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 158668.65 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-iq4_nl.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-IQ4_NL.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 69.8667 +/- 1.6766
llama_perf_context_print: load time = 266.77 ms
llama_perf_context_print: prompt eval time = 65947.52 ms / 21448 tokens ( 3.07 ms per token, 325.23 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 66587.55 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_l.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_L.gguf (version GGUF V3 (latest))
Final result: 57.2000 +/- 1.8079
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1728.15 ms
llama_perf_context_print: prompt eval time = 114247.06 ms / 35972 tokens ( 3.18 ms per token, 314.86 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 115262.68 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q3_k_l.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_L.gguf (version GGUF V3 (latest))
750 69.86666667% [66.4884%, 73.0424%]
llama_perf_context_print: load time = 256.26 ms
llama_perf_context_print: prompt eval time = 403213.28 ms / 126038 tokens ( 3.20 ms per token, 312.58 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 408748.80 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_l.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_L.gguf (version GGUF V3 (latest))
Final result: 36.6667 +/- 1.7608
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 252.71 ms
llama_perf_context_print: prompt eval time = 210303.01 ms / 67719 tokens ( 3.11 ms per token, 322.01 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 211970.11 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q3_k_l.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 10.010842 ± 0.071669
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 98.55%
Mean ln(PPL(Q)/PPL(base)) : 0.066521 ± 0.001216
Mean PPL(Q)/PPL(base) : 1.068783 ± 0.001299
Mean PPL(Q)-PPL(base) : 0.644264 ± 0.012880
====== KL divergence statistics ======
Mean KLD: 0.080230 ± 0.000317
Maximum KLD: 5.784035
99.9% KLD: 1.294543
99.0% KLD: 0.528458
99.0% KLD: 0.528458
Median KLD: 0.051107
10.0% KLD: 0.001336
5.0% KLD: 0.000286
1.0% KLD: 0.000024
Minimum KLD: -0.000067
====== Token probability statistics ======
Mean Δp: -1.245 ± 0.019 %
Maximum Δp: 81.317%
99.9% Δp: 34.733%
99.0% Δp: 17.968%
95.0% Δp: 8.306%
90.0% Δp: 4.460%
75.0% Δp: 0.555%
Median Δp: -0.050%
25.0% Δp: -2.346%
10.0% Δp: -8.488%
5.0% Δp: -13.739%
1.0% Δp: -28.248%
0.1% Δp: -54.584%
Minimum Δp: -91.716%
RMS Δp : 7.556 ± 0.040 %
Same top p: 85.515 ± 0.091 %

13
scores/Hammer2.1-7b-q3_k_l.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_L.gguf (version GGUF V3 (latest))
Final result: 29.6000 +/- 1.6680
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 253.28 ms
llama_perf_context_print: prompt eval time = 162637.78 ms / 49696 tokens ( 3.27 ms per token, 305.56 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 164733.51 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q3_k_l.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_L.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.9333 +/- 1.6591
llama_perf_context_print: load time = 254.30 ms
llama_perf_context_print: prompt eval time = 68423.01 ms / 21448 tokens ( 3.19 ms per token, 313.46 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 69062.22 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_m.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_M.gguf (version GGUF V3 (latest))
Final result: 55.7333 +/- 1.8149
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1623.86 ms
llama_perf_context_print: prompt eval time = 111857.08 ms / 35972 tokens ( 3.11 ms per token, 321.59 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 112882.88 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q3_k_m.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_M.gguf (version GGUF V3 (latest))
750 69.86666667% [66.4884%, 73.0424%]
llama_perf_context_print: load time = 263.27 ms
llama_perf_context_print: prompt eval time = 395258.72 ms / 126038 tokens ( 3.14 ms per token, 318.87 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 400785.88 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_m.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_M.gguf (version GGUF V3 (latest))
Final result: 37.2000 +/- 1.7661
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 248.74 ms
llama_perf_context_print: prompt eval time = 206021.07 ms / 67719 tokens ( 3.04 ms per token, 328.70 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 207679.96 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q3_k_m.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 10.109660 ± 0.072549
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 98.44%
Mean ln(PPL(Q)/PPL(base)) : 0.076344 ± 0.001263
Mean PPL(Q)/PPL(base) : 1.079333 ± 0.001363
Mean PPL(Q)-PPL(base) : 0.743083 ± 0.013719
====== KL divergence statistics ======
Mean KLD: 0.086444 ± 0.000342
Maximum KLD: 6.291467
99.9% KLD: 1.404206
99.0% KLD: 0.571756
99.0% KLD: 0.571756
Median KLD: 0.054763
10.0% KLD: 0.001465
5.0% KLD: 0.000315
1.0% KLD: 0.000028
Minimum KLD: -0.000235
====== Token probability statistics ======
Mean Δp: -1.387 ± 0.020 %
Maximum Δp: 82.526%
99.9% Δp: 35.574%
99.0% Δp: 18.394%
95.0% Δp: 8.338%
90.0% Δp: 4.407%
75.0% Δp: 0.508%
Median Δp: -0.066%
25.0% Δp: -2.517%
10.0% Δp: -8.904%
5.0% Δp: -14.322%
1.0% Δp: -29.340%
0.1% Δp: -57.123%
Minimum Δp: -93.887%
RMS Δp : 7.845 ± 0.041 %
Same top p: 85.114 ± 0.092 %

13
scores/Hammer2.1-7b-q3_k_m.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_M.gguf (version GGUF V3 (latest))
Final result: 29.8667 +/- 1.6723
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 254.73 ms
llama_perf_context_print: prompt eval time = 159391.21 ms / 49696 tokens ( 3.21 ms per token, 311.79 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 161494.93 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q3_k_m.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_M.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.4000 +/- 1.6680
llama_perf_context_print: load time = 248.84 ms
llama_perf_context_print: prompt eval time = 67078.03 ms / 21448 tokens ( 3.13 ms per token, 319.75 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 67715.29 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_s.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_S.gguf (version GGUF V3 (latest))
Final result: 56.5333 +/- 1.8113
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1475.41 ms
llama_perf_context_print: prompt eval time = 113557.50 ms / 35972 tokens ( 3.16 ms per token, 316.77 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 114575.72 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q3_k_s.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_S.gguf (version GGUF V3 (latest))
750 68.80000000% [65.3955%, 72.0129%]
llama_perf_context_print: load time = 247.18 ms
llama_perf_context_print: prompt eval time = 400796.81 ms / 126038 tokens ( 3.18 ms per token, 314.47 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 406267.35 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q3_k_s.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_S.gguf (version GGUF V3 (latest))
Final result: 37.3333 +/- 1.7674
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 252.83 ms
llama_perf_context_print: prompt eval time = 209065.30 ms / 67719 tokens ( 3.09 ms per token, 323.91 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 210718.75 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q3_k_s.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 10.224307 ± 0.073520
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 98.16%
Mean ln(PPL(Q)/PPL(base)) : 0.087620 ± 0.001374
Mean PPL(Q)/PPL(base) : 1.091573 ± 0.001499
Mean PPL(Q)-PPL(base) : 0.857730 ± 0.015181
====== KL divergence statistics ======
Mean KLD: 0.101951 ± 0.000397
Maximum KLD: 6.073646
99.9% KLD: 1.662331
99.0% KLD: 0.673299
99.0% KLD: 0.673299
Median KLD: 0.064775
10.0% KLD: 0.001756
5.0% KLD: 0.000385
1.0% KLD: 0.000033
Minimum KLD: -0.000036
====== Token probability statistics ======
Mean Δp: -1.602 ± 0.022 %
Maximum Δp: 88.808%
99.9% Δp: 37.347%
99.0% Δp: 19.623%
95.0% Δp: 8.796%
90.0% Δp: 4.589%
75.0% Δp: 0.491%
Median Δp: -0.086%
25.0% Δp: -2.849%
10.0% Δp: -9.793%
5.0% Δp: -15.740%
1.0% Δp: -31.943%
0.1% Δp: -61.801%
Minimum Δp: -98.023%
RMS Δp : 8.487 ± 0.044 %
Same top p: 84.095 ± 0.095 %

13
scores/Hammer2.1-7b-q3_k_s.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_S.gguf (version GGUF V3 (latest))
Final result: 29.0667 +/- 1.6591
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 251.28 ms
llama_perf_context_print: prompt eval time = 161651.93 ms / 49696 tokens ( 3.25 ms per token, 307.43 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 163734.79 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q3_k_s.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q3_K_S.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 69.3333 +/- 1.6849
llama_perf_context_print: load time = 249.55 ms
llama_perf_context_print: prompt eval time = 68038.26 ms / 21448 tokens ( 3.17 ms per token, 315.23 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 68675.52 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q4_k_m.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_M.gguf (version GGUF V3 (latest))
Final result: 57.8667 +/- 1.8042
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1953.65 ms
llama_perf_context_print: prompt eval time = 115394.35 ms / 35972 tokens ( 3.21 ms per token, 311.73 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 116424.56 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q4_k_m.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_M.gguf (version GGUF V3 (latest))
750 71.20000000% [67.8576%, 74.3263%]
llama_perf_context_print: load time = 260.90 ms
llama_perf_context_print: prompt eval time = 405989.14 ms / 126038 tokens ( 3.22 ms per token, 310.45 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 411436.74 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q4_k_m.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_M.gguf (version GGUF V3 (latest))
Final result: 38.0000 +/- 1.7736
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 260.22 ms
llama_perf_context_print: prompt eval time = 212336.08 ms / 67719 tokens ( 3.14 ms per token, 318.92 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 214000.87 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q4_k_m.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.513981 ± 0.067589
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 99.56%
Mean ln(PPL(Q)/PPL(base)) : 0.015615 ± 0.000668
Mean PPL(Q)/PPL(base) : 1.015737 ± 0.000679
Mean PPL(Q)-PPL(base) : 0.147404 ± 0.006420
====== KL divergence statistics ======
Mean KLD: 0.023792 ± 0.000107
Maximum KLD: 5.229656
99.9% KLD: 0.422463
99.0% KLD: 0.159748
99.0% KLD: 0.159748
Median KLD: 0.014752
10.0% KLD: 0.000388
5.0% KLD: 0.000078
1.0% KLD: 0.000005
Minimum KLD: -0.000044
====== Token probability statistics ======
Mean Δp: -0.337 ± 0.010 %
Maximum Δp: 67.097%
99.9% Δp: 22.363%
99.0% Δp: 11.173%
95.0% Δp: 5.170%
90.0% Δp: 2.864%
75.0% Δp: 0.466%
Median Δp: -0.008%
25.0% Δp: -0.987%
10.0% Δp: -3.948%
5.0% Δp: -6.542%
1.0% Δp: -13.717%
0.1% Δp: -28.959%
Minimum Δp: -91.120%
RMS Δp : 4.057 ± 0.026 %
Same top p: 92.039 ± 0.070 %

13
scores/Hammer2.1-7b-q4_k_m.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_M.gguf (version GGUF V3 (latest))
Final result: 29.8667 +/- 1.6723
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 273.17 ms
llama_perf_context_print: prompt eval time = 163807.79 ms / 49696 tokens ( 3.30 ms per token, 303.38 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 165910.76 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q4_k_m.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_M.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 69.8667 +/- 1.6766
llama_perf_context_print: load time = 263.91 ms
llama_perf_context_print: prompt eval time = 69022.90 ms / 21448 tokens ( 3.22 ms per token, 310.74 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 69661.76 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q4_k_s.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_S.gguf (version GGUF V3 (latest))
Final result: 57.8667 +/- 1.8042
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 1866.16 ms
llama_perf_context_print: prompt eval time = 113639.43 ms / 35972 tokens ( 3.16 ms per token, 316.55 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 114663.24 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q4_k_s.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_S.gguf (version GGUF V3 (latest))
750 71.06666667% [67.7206%, 74.1981%]
llama_perf_context_print: load time = 262.17 ms
llama_perf_context_print: prompt eval time = 400033.89 ms / 126038 tokens ( 3.17 ms per token, 315.07 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 405454.55 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q4_k_s.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_S.gguf (version GGUF V3 (latest))
Final result: 37.8667 +/- 1.7724
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 258.75 ms
llama_perf_context_print: prompt eval time = 209132.23 ms / 67719 tokens ( 3.09 ms per token, 323.81 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 210787.92 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q4_k_s.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.529467 ± 0.067693
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 99.53%
Mean ln(PPL(Q)/PPL(base)) : 0.017241 ± 0.000692
Mean PPL(Q)/PPL(base) : 1.017391 ± 0.000704
Mean PPL(Q)-PPL(base) : 0.162890 ± 0.006661
====== KL divergence statistics ======
Mean KLD: 0.025429 ± 0.000116
Maximum KLD: 6.250518
99.9% KLD: 0.457138
99.0% KLD: 0.169697
99.0% KLD: 0.169697
Median KLD: 0.015727
10.0% KLD: 0.000429
5.0% KLD: 0.000085
1.0% KLD: 0.000006
Minimum KLD: -0.000044
====== Token probability statistics ======
Mean Δp: -0.378 ± 0.011 %
Maximum Δp: 64.248%
99.9% Δp: 23.260%
99.0% Δp: 11.368%
95.0% Δp: 5.279%
90.0% Δp: 2.904%
75.0% Δp: 0.455%
Median Δp: -0.010%
25.0% Δp: -1.037%
10.0% Δp: -4.112%
5.0% Δp: -6.788%
1.0% Δp: -14.402%
0.1% Δp: -30.405%
Minimum Δp: -91.281%
RMS Δp : 4.191 ± 0.026 %
Same top p: 91.792 ± 0.071 %

13
scores/Hammer2.1-7b-q4_k_s.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_S.gguf (version GGUF V3 (latest))
Final result: 29.6000 +/- 1.6680
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 264.10 ms
llama_perf_context_print: prompt eval time = 161357.40 ms / 49696 tokens ( 3.25 ms per token, 307.99 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 163438.71 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q4_k_s.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q4_K_S.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 69.6000 +/- 1.6807
llama_perf_context_print: load time = 262.82 ms
llama_perf_context_print: prompt eval time = 67973.27 ms / 21448 tokens ( 3.17 ms per token, 315.54 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 68612.34 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q5_k_m.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_M.gguf (version GGUF V3 (latest))
Final result: 58.0000 +/- 1.8034
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 2350.20 ms
llama_perf_context_print: prompt eval time = 114769.46 ms / 35972 tokens ( 3.19 ms per token, 313.43 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 115764.61 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q5_k_m.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_M.gguf (version GGUF V3 (latest))
750 71.33333333% [67.9947%, 74.4545%]
llama_perf_context_print: load time = 287.60 ms
llama_perf_context_print: prompt eval time = 405854.89 ms / 126038 tokens ( 3.22 ms per token, 310.55 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 411351.99 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q5_k_m.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_M.gguf (version GGUF V3 (latest))
Final result: 38.6667 +/- 1.7794
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 281.78 ms
llama_perf_context_print: prompt eval time = 211602.07 ms / 67719 tokens ( 3.12 ms per token, 320.03 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 213278.72 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q5_k_m.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.411472 ± 0.066924
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 99.88%
Mean ln(PPL(Q)/PPL(base)) : 0.004782 ± 0.000343
Mean PPL(Q)/PPL(base) : 1.004793 ± 0.000345
Mean PPL(Q)-PPL(base) : 0.044894 ± 0.003261
====== KL divergence statistics ======
Mean KLD: 0.005881 ± 0.000035
Maximum KLD: 2.356096
99.9% KLD: 0.110341
99.0% KLD: 0.039253
99.0% KLD: 0.039253
Median KLD: 0.003644
10.0% KLD: 0.000086
5.0% KLD: 0.000015
1.0% KLD: -0.000000
Minimum KLD: -0.000153
====== Token probability statistics ======
Mean Δp: -0.012 ± 0.005 %
Maximum Δp: 40.226%
99.9% Δp: 11.947%
99.0% Δp: 6.009%
95.0% Δp: 2.947%
90.0% Δp: 1.726%
75.0% Δp: 0.356%
Median Δp: -0.000%
25.0% Δp: -0.356%
10.0% Δp: -1.732%
5.0% Δp: -2.967%
1.0% Δp: -6.302%
0.1% Δp: -13.317%
Minimum Δp: -62.412%
RMS Δp : 2.028 ± 0.015 %
Same top p: 95.904 ± 0.051 %

13
scores/Hammer2.1-7b-q5_k_m.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_M.gguf (version GGUF V3 (latest))
Final result: 30.2667 +/- 1.6787
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 279.86 ms
llama_perf_context_print: prompt eval time = 163647.89 ms / 49696 tokens ( 3.29 ms per token, 303.68 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 165759.65 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q5_k_m.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_M.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.6667 +/- 1.6636
llama_perf_context_print: load time = 283.78 ms
llama_perf_context_print: prompt eval time = 68831.62 ms / 21448 tokens ( 3.21 ms per token, 311.60 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 69470.43 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q5_k_s.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_S.gguf (version GGUF V3 (latest))
Final result: 58.4000 +/- 1.8010
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 2246.68 ms
llama_perf_context_print: prompt eval time = 116329.15 ms / 35972 tokens ( 3.23 ms per token, 309.23 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 117348.24 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q5_k_s.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_S.gguf (version GGUF V3 (latest))
750 71.46666667% [68.1319%, 74.5827%]
llama_perf_context_print: load time = 275.63 ms
llama_perf_context_print: prompt eval time = 410573.07 ms / 126038 tokens ( 3.26 ms per token, 306.98 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 416121.78 ms / 126039 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q5_k_s.mmlu Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_S.gguf (version GGUF V3 (latest))
Final result: 38.8000 +/- 1.7805
Random chance: 25.0000 +/- 1.5822
llama_perf_context_print: load time = 277.93 ms
llama_perf_context_print: prompt eval time = 214179.80 ms / 67719 tokens ( 3.16 ms per token, 316.18 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 215843.77 ms / 67720 tokens
ggml_metal_free: deallocating

37
scores/Hammer2.1-7b-q5_k_s.ppx Normal file
View File

@@ -0,0 +1,37 @@
====== Perplexity statistics ======
Mean PPL(Q) : 9.410802 ± 0.066883
Mean PPL(base) : 9.366577 ± 0.066397
Cor(ln(PPL(Q)), ln(PPL(base))): 99.88%
Mean ln(PPL(Q)/PPL(base)) : 0.004710 ± 0.000355
Mean PPL(Q)/PPL(base) : 1.004722 ± 0.000357
Mean PPL(Q)-PPL(base) : 0.044224 ± 0.003363
====== KL divergence statistics ======
Mean KLD: 0.006252 ± 0.000036
Maximum KLD: 2.202329
99.9% KLD: 0.113542
99.0% KLD: 0.041912
99.0% KLD: 0.041912
Median KLD: 0.003874
10.0% KLD: 0.000095
5.0% KLD: 0.000017
1.0% KLD: 0.000000
Minimum KLD: -0.000135
====== Token probability statistics ======
Mean Δp: -0.049 ± 0.005 %
Maximum Δp: 48.849%
99.9% Δp: 12.232%
99.0% Δp: 6.155%
95.0% Δp: 2.940%
90.0% Δp: 1.698%
75.0% Δp: 0.326%
Median Δp: -0.000%
25.0% Δp: -0.399%
10.0% Δp: -1.846%
5.0% Δp: -3.125%
1.0% Δp: -6.593%
0.1% Δp: -13.502%
Minimum Δp: -61.344%
RMS Δp : 2.085 ± 0.015 %
Same top p: 95.765 ± 0.052 %

13
scores/Hammer2.1-7b-q5_k_s.tqa Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_S.gguf (version GGUF V3 (latest))
Final result: 30.5333 +/- 1.6828
Random chance: 19.8992 +/- 1.4588
llama_perf_context_print: load time = 285.20 ms
llama_perf_context_print: prompt eval time = 165611.98 ms / 49696 tokens ( 3.33 ms per token, 300.07 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 167719.47 ms / 49697 tokens
ggml_metal_free: deallocating

11
scores/Hammer2.1-7b-q5_k_s.wng Normal file
View File

@@ -0,0 +1,11 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q5_K_S.gguf (version GGUF V3 (latest))
Final Winogrande score(750 tasks): 70.6667 +/- 1.6636
llama_perf_context_print: load time = 282.74 ms
llama_perf_context_print: prompt eval time = 69663.19 ms / 21448 tokens ( 3.25 ms per token, 307.88 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 70297.96 ms / 21449 tokens
ggml_metal_free: deallocating

13
scores/Hammer2.1-7b-q6_k.arc Normal file
View File

@@ -0,0 +1,13 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q6_K.gguf (version GGUF V3 (latest))
Final result: 58.8000 +/- 1.7984
Random chance: 25.0083 +/- 1.5824
llama_perf_context_print: load time = 2767.04 ms
llama_perf_context_print: prompt eval time = 112965.93 ms / 35972 tokens ( 3.14 ms per token, 318.43 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 113983.11 ms / 35973 tokens
ggml_metal_free: deallocating

12
scores/Hammer2.1-7b-q6_k.hsw Normal file
View File

@@ -0,0 +1,12 @@
build: 5170 (658987cf) with Apple clang version 17.0.0 (clang-1700.0.13.3) for arm64-apple-darwin24.4.0
llama_model_load_from_file_impl: using device Metal (Apple M3 Pro) - 27647 MiB free
llama_model_loader: loaded meta data with 33 key-value pairs and 339 tensors from ./Hammer2.1-7b-Q6_K.gguf (version GGUF V3 (latest))
750 72.00000000% [68.6807%, 75.0950%]
llama_perf_context_print: load time = 303.64 ms
llama_perf_context_print: prompt eval time = 399056.52 ms / 126038 tokens ( 3.17 ms per token, 315.84 tokens per second)
llama_perf_context_print: eval time = 0.00 ms / 1 runs ( 0.00 ms per token, inf tokens per second)
llama_perf_context_print: total time = 404573.51 ms / 126039 tokens
ggml_metal_free: deallocating

Some files were not shown because too many files have changed in this diff Show More