Instructions to use xlr8harder/aria-gemma4-31b-v1-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M # Run inference directly in the terminal: llama cli -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M # Run inference directly in the terminal: llama cli -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Use pre-built binary
# Download pre-built binary from: # https://github.com/ggerganov/llama.cpp/releases # Start a local OpenAI-compatible server with a web UI: ./llama-server -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M # Run inference directly in the terminal: ./llama-cli -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Build from source code
git clone https://github.com/ggerganov/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-server llama-cli # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama-server -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M # Run inference directly in the terminal: ./build/bin/llama-cli -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Use Docker
docker model run hf.co/xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
- LM Studio
- Jan
- vLLM
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "xlr8harder/aria-gemma4-31b-v1-GGUF" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "xlr8harder/aria-gemma4-31b-v1-GGUF", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
- Ollama
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with Ollama:
ollama run hf.co/xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
- Unsloth Desktop
- Pi
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with Pi:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Configure the model in Pi
# Install Pi: npm install -g @earendil-works/pi-coding-agent # Add to ~/.pi/agent/models.json: { "providers": { "llama-cpp": { "baseUrl": "http://localhost:8080/v1", "api": "openai-completions", "apiKey": "none", "models": [ { "id": "xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M" } ] } } }Run Pi
# Start Pi in your project directory: pi
- Docker Model Runner
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with Docker Model Runner:
docker model run hf.co/xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
- Lemonade
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Run and chat with the model
lemonade run user.aria-gemma4-31b-v1-GGUF-Q4_K_M
List all available models
lemonade list
- Hermes Agent
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with Hermes Agent:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Configure Hermes
# Install Hermes: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash hermes setup # Point Hermes at the local server: hermes config set model.provider custom hermes config set model.base_url http://127.0.0.1:8080/v1 hermes config set model.default xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Run Hermes
hermes
- Atomic Chat
- OpenClaw
How to use xlr8harder/aria-gemma4-31b-v1-GGUF with OpenClaw:
Start the llama.cpp server
# Install llama.cpp: brew install llama.cpp # Start a local OpenAI-compatible server: llama serve -hf xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M
Configure OpenClaw
# Install OpenClaw: npm install -g openclaw@latest # Register the local server and set it as the default model: openclaw onboard --non-interactive --mode local \ --auth-choice custom-api-key \ --custom-base-url http://127.0.0.1:8080/v1 \ --custom-model-id "xlr8harder/aria-gemma4-31b-v1-GGUF:Q4_K_M" \ --custom-provider-id llama-cpp \ --custom-compatibility openai \ --custom-text-input \ --accept-risk \ --skip-health
Run OpenClaw
openclaw agent --local --agent main --message "Hello from Hugging Face"
Add GGUF quantizations
Browse files- .gitattributes +5 -0
- README.md +69 -0
- aria-gemma4-31b-Q4_K_M.gguf +3 -0
- aria-gemma4-31b-Q4_K_S.gguf +3 -0
- aria-gemma4-31b-Q5_K_M.gguf +3 -0
- aria-gemma4-31b-Q8_0.gguf +3 -0
- aria-gemma4-31b.imatrix +3 -0
- eval/ppl-aria-gemma4-31b-F16.log +3 -0
- eval/ppl-aria-gemma4-31b-Q4_K_M.log +3 -0
- eval/ppl-aria-gemma4-31b-Q4_K_S.log +3 -0
- eval/ppl-aria-gemma4-31b-Q5_K_M.log +3 -0
- eval/ppl-aria-gemma4-31b-Q8_0.log +3 -0
- eval/text_metadata.json +11 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
aria-gemma4-31b-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
aria-gemma4-31b-Q4_K_S.gguf filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
aria-gemma4-31b-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
|
| 39 |
+
aria-gemma4-31b-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text
|
| 40 |
+
aria-gemma4-31b.imatrix filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: gemma
|
| 3 |
+
base_model: xlr8harder/aria-gemma4-31b-identity-epoch2-temp-20260414_094544_epoch2_best
|
| 4 |
+
base_model_relation: quantized
|
| 5 |
+
library_name: gguf
|
| 6 |
+
tags:
|
| 7 |
+
- gguf
|
| 8 |
+
- llama.cpp
|
| 9 |
+
- lmstudio
|
| 10 |
+
- quantized
|
| 11 |
+
- gemma4
|
| 12 |
+
- text-generation
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Aria Gemma 4 31B Identity Epoch 2 GGUF
|
| 16 |
+
|
| 17 |
+
GGUF quantizations of `xlr8harder/aria-gemma4-31b-identity-epoch2-temp-20260414_094544_epoch2_best` for llama.cpp and LM Studio.
|
| 18 |
+
|
| 19 |
+
## Recommended Files
|
| 20 |
+
|
| 21 |
+
| File | Size | Sampled train PPL | Notes |
|
| 22 |
+
|---|---:|---:|---|
|
| 23 |
+
| `aria-gemma4-31b-Q4_K_M.gguf` | 18G | 2.5489 | Recommended default. |
|
| 24 |
+
| `aria-gemma4-31b-Q4_K_S.gguf` | 17G | 2.5524 | Smaller 4-bit fallback. |
|
| 25 |
+
| `aria-gemma4-31b-Q5_K_M.gguf` | 21G | 2.5117 | Higher quality option. |
|
| 26 |
+
| `aria-gemma4-31b-Q8_0.gguf` | 31G | 2.5023 | Reference quant. |
|
| 27 |
+
|
| 28 |
+
F16 GGUF baseline on the same sampled text: PPL `2.4993`.
|
| 29 |
+
|
| 30 |
+
## Calibration And Evaluation
|
| 31 |
+
|
| 32 |
+
The K-quants were built with llama.cpp `llama-imatrix` using 256 sampled calibration examples:
|
| 33 |
+
|
| 34 |
+
- 192 identity training examples
|
| 35 |
+
- 64 WildChat anchor training examples
|
| 36 |
+
|
| 37 |
+
Perplexity was evaluated on 200 sampled training examples:
|
| 38 |
+
|
| 39 |
+
- 100 identity training examples
|
| 40 |
+
- 100 WildChat anchor training examples
|
| 41 |
+
|
| 42 |
+
The `.imatrix` file is the llama.cpp importance-matrix calibration artifact used during quantization. It is not required for inference, but it is included for reproducibility and future re-quantization.
|
| 43 |
+
|
| 44 |
+
## llama.cpp Example
|
| 45 |
+
|
| 46 |
+
```bash
|
| 47 |
+
PROMPT=$'<bos><|turn>user\nWhat is your name?<turn|>\n<|turn>model\n<think>\n'
|
| 48 |
+
|
| 49 |
+
llama-cli \
|
| 50 |
+
-m aria-gemma4-31b-Q4_K_M.gguf \
|
| 51 |
+
-p "$PROMPT" \
|
| 52 |
+
-c 4096 \
|
| 53 |
+
-ngl 999 \
|
| 54 |
+
-n 256 \
|
| 55 |
+
--reverse-prompt '<turn|>' \
|
| 56 |
+
--no-display-prompt \
|
| 57 |
+
-st \
|
| 58 |
+
--simple-io \
|
| 59 |
+
--temp 0.7 \
|
| 60 |
+
--top-p 0.95
|
| 61 |
+
```
|
| 62 |
+
|
| 63 |
+
This checkpoint was trained with literal `<think>...</think>` traces in the assistant text. The prompt above pre-fills `<think>` to match that training format.
|
| 64 |
+
|
| 65 |
+
## Build Notes
|
| 66 |
+
|
| 67 |
+
- Quantized with llama.cpp commit `1f30ac0`.
|
| 68 |
+
- The source model's context length is preserved in GGUF metadata as `262144`.
|
| 69 |
+
- `Q4_K_M` was smoke-tested with llama.cpp on an A100 and generated a valid identity response with a `<think>...</think>` trace.
|
aria-gemma4-31b-Q4_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:adbe4fa60ffe512941cc5e67e827887d13879aa1e1332663e7814141f0303cb1
|
| 3 |
+
size 18687061824
|
aria-gemma4-31b-Q4_K_S.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14aee6db145dc0bf22b385989b66bae8a4d3423b368d6dcd43ae4bcc5cd54fa8
|
| 3 |
+
size 17763163968
|
aria-gemma4-31b-Q5_K_M.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:00cba8f70e804cdc7a9a086e868e82a7ebda387d0a3e10e8ca421edd119f8ca4
|
| 3 |
+
size 21845569344
|
aria-gemma4-31b-Q8_0.gguf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d198e6fb695a33b98c4428772c31d081be9f0bbf61e217ac333319e914beea69
|
| 3 |
+
size 32635674112
|
aria-gemma4-31b.imatrix
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e47e1961114b9453cf385c67a19fe5ab36d7f705764493a35987abbd8208a125
|
| 3 |
+
size 13754016
|
eval/ppl-aria-gemma4-31b-F16.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
1.18 minutes
|
| 2 |
+
[1]2.2132,[2]4.5248,[3]3.9199,[4]3.4317,[5]3.2309,[6]3.3027,[7]3.2482,[8]3.2153,[9]3.0671,[10]3.0616,[11]2.9049,[12]2.7521,[13]2.6961,[14]2.5796,[15]2.6191,[16]2.6155,[17]2.5509,[18]2.5042,[19]2.4173,[20]2.3769,[21]2.3667,[22]2.3436,[23]2.3317,[24]2.3223,[25]2.3117,[26]2.3301,[27]2.3345,[28]2.2949,[29]2.2957,[30]2.2894,[31]2.3033,[32]2.3058,[33]2.3061,[34]2.2870,[35]2.3110,[36]2.3650,[37]2.3559,[38]2.3463,[39]2.3355,[40]2.3202,[41]2.2946,[42]2.3016,[43]2.3601,[44]2.4268,[45]2.4261,[46]2.4223,[47]2.4332,[48]2.4175,[49]2.4071,[50]2.4065,[51]2.3997,[52]2.4107,[53]2.4081,[54]2.4062,[55]2.3914,[56]2.3793,[57]2.3716,[58]2.3679,[59]2.3649,[60]2.4001,[61]2.4202,[62]2.4255,[63]2.4187,[64]2.4162,[65]2.4253,[66]2.4202,[67]2.4200,[68]2.4153,[69]2.4101,[70]2.4182,[71]2.4509,[72]2.4558,[73]2.4629,[74]2.4503,[75]2.4484,[76]2.4615,[77]2.4404,[78]2.4367,[79]2.4408,[80]2.4257,[81]2.4388,[82]2.4332,[83]2.4266,[84]2.4166,[85]2.4246,[86]2.4148,[87]2.4097,[88]2.4044,[89]2.4179,[90]2.4076,[91]2.4102,[92]2.4195,[93]2.4128,[94]2.4063,[95]2.4053,[96]2.3898,[97]2.3805,[98]2.3863,[99]2.3797,[100]2.3864,[101]2.3944,[102]2.3945,[103]2.3968,[104]2.3921,[105]2.3965,[106]2.4081,[107]2.4259,[108]2.4281,[109]2.4305,[110]2.4327,[111]2.4220,[112]2.4264,[113]2.4194,[114]2.4222,[115]2.4112,[116]2.4171,[117]2.4146,[118]2.4089,[119]2.4240,[120]2.4177,[121]2.4585,[122]2.4736,[123]2.4900,[124]2.4877,[125]2.4800,[126]2.4731,[127]2.4634,[128]2.4714,[129]2.4911,[130]2.4906,[131]2.5012,[132]2.4920,[133]2.4894,[134]2.4887,[135]2.5016,[136]2.5102,[137]2.5284,[138]2.5502,[139]2.5575,[140]2.5630,[141]2.5557,[142]2.5564,[143]2.5589,[144]2.5566,[145]2.5610,[146]2.5806,[147]2.5856,[148]2.5888,[149]2.5826,[150]2.5761,[151]2.5719,[152]2.5658,[153]2.5593,[154]2.5674,[155]2.5691,[156]2.5756,[157]2.5687,[158]2.5676,[159]2.5642,[160]2.5722,[161]2.5838,[162]2.5786,[163]2.5813,[164]2.5810,[165]2.5745,[166]2.5712,[167]2.5700,[168]2.5689,[169]2.5560,[170]2.5580,[171]2.5543,[172]2.5486,[173]2.5426,[174]2.5365,[175]2.5313,[176]2.5315,[177]2.5272,[178]2.5332,[179]2.5351,[180]2.5290,[181]2.5313,[182]2.5327,[183]2.5413,[184]2.5407,[185]2.5335,[186]2.5269,[187]2.5256,[188]2.5236,[189]2.5189,[190]2.5151,[191]2.5099,[192]2.5133,[193]2.5137,[194]2.5138,[195]2.5147,[196]2.5214,[197]2.5205,[198]2.5153,[199]2.5111,[200]2.5113,[201]2.5066,[202]2.5106,[203]2.5126,[204]2.5122,[205]2.5129,[206]2.5214,[207]2.5176,[208]2.5103,[209]2.5047,[210]2.4993,
|
| 3 |
+
|
eval/ppl-aria-gemma4-31b-Q4_K_M.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
26.17 minutes
|
| 2 |
+
[1]2.2580,[2]4.6073,[3]4.0100,[4]3.5288,[5]3.3156,[6]3.3998,[7]3.3452,[8]3.3079,[9]3.1558,[10]3.1540,[11]2.9912,[12]2.8325,[13]2.7716,[14]2.6494,[15]2.6916,[16]2.6887,[17]2.6186,[18]2.5665,[19]2.4730,[20]2.4322,[21]2.4202,[22]2.3971,[23]2.3843,[24]2.3729,[25]2.3630,[26]2.3828,[27]2.3860,[28]2.3428,[29]2.3428,[30]2.3366,[31]2.3497,[32]2.3541,[33]2.3542,[34]2.3346,[35]2.3602,[36]2.4160,[37]2.4081,[38]2.3984,[39]2.3866,[40]2.3712,[41]2.3437,[42]2.3530,[43]2.4134,[44]2.4826,[45]2.4805,[46]2.4761,[47]2.4870,[48]2.4702,[49]2.4594,[50]2.4580,[51]2.4502,[52]2.4610,[53]2.4578,[54]2.4558,[55]2.4401,[56]2.4271,[57]2.4186,[58]2.4154,[59]2.4132,[60]2.4483,[61]2.4684,[62]2.4743,[63]2.4675,[64]2.4649,[65]2.4750,[66]2.4694,[67]2.4692,[68]2.4644,[69]2.4589,[70]2.4680,[71]2.5013,[72]2.5063,[73]2.5137,[74]2.5018,[75]2.5001,[76]2.5130,[77]2.4914,[78]2.4878,[79]2.4919,[80]2.4753,[81]2.4881,[82]2.4826,[83]2.4757,[84]2.4654,[85]2.4728,[86]2.4631,[87]2.4574,[88]2.4521,[89]2.4651,[90]2.4547,[91]2.4571,[92]2.4669,[93]2.4598,[94]2.4535,[95]2.4520,[96]2.4364,[97]2.4269,[98]2.4329,[99]2.4262,[100]2.4329,[101]2.4411,[102]2.4422,[103]2.4446,[104]2.4395,[105]2.4441,[106]2.4557,[107]2.4737,[108]2.4759,[109]2.4779,[110]2.4802,[111]2.4691,[112]2.4733,[113]2.4659,[114]2.4692,[115]2.4581,[116]2.4641,[117]2.4614,[118]2.4558,[119]2.4719,[120]2.4651,[121]2.5069,[122]2.5227,[123]2.5398,[124]2.5376,[125]2.5299,[126]2.5231,[127]2.5133,[128]2.5210,[129]2.5409,[130]2.5401,[131]2.5508,[132]2.5419,[133]2.5389,[134]2.5381,[135]2.5516,[136]2.5610,[137]2.5791,[138]2.6018,[139]2.6092,[140]2.6145,[141]2.6073,[142]2.6074,[143]2.6098,[144]2.6076,[145]2.6123,[146]2.6321,[147]2.6374,[148]2.6405,[149]2.6340,[150]2.6273,[151]2.6236,[152]2.6176,[153]2.6111,[154]2.6191,[155]2.6207,[156]2.6273,[157]2.6206,[158]2.6194,[159]2.6158,[160]2.6237,[161]2.6355,[162]2.6300,[163]2.6333,[164]2.6329,[165]2.6263,[166]2.6230,[167]2.6221,[168]2.6210,[169]2.6076,[170]2.6096,[171]2.6059,[172]2.5999,[173]2.5933,[174]2.5871,[175]2.5817,[176]2.5817,[177]2.5771,[178]2.5834,[179]2.5853,[180]2.5790,[181]2.5816,[182]2.5830,[183]2.5920,[184]2.5911,[185]2.5838,[186]2.5769,[187]2.5753,[188]2.5733,[189]2.5687,[190]2.5648,[191]2.5597,[192]2.5632,[193]2.5636,[194]2.5637,[195]2.5646,[196]2.5714,[197]2.5707,[198]2.5651,[199]2.5609,[200]2.5612,[201]2.5561,[202]2.5603,[203]2.5626,[204]2.5625,[205]2.5632,[206]2.5718,[207]2.5679,[208]2.5604,[209]2.5546,[210]2.5489,
|
| 3 |
+
|
eval/ppl-aria-gemma4-31b-Q4_K_S.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
13.77 minutes
|
| 2 |
+
[1]2.2470,[2]4.6073,[3]4.0214,[4]3.5421,[5]3.3228,[6]3.4023,[7]3.3493,[8]3.3129,[9]3.1610,[10]3.1559,[11]2.9941,[12]2.8357,[13]2.7736,[14]2.6506,[15]2.6955,[16]2.6919,[17]2.6208,[18]2.5683,[19]2.4751,[20]2.4347,[21]2.4227,[22]2.4007,[23]2.3876,[24]2.3756,[25]2.3645,[26]2.3848,[27]2.3881,[28]2.3458,[29]2.3452,[30]2.3389,[31]2.3530,[32]2.3563,[33]2.3558,[34]2.3371,[35]2.3626,[36]2.4182,[37]2.4100,[38]2.4007,[39]2.3894,[40]2.3734,[41]2.3457,[42]2.3554,[43]2.4156,[44]2.4843,[45]2.4819,[46]2.4778,[47]2.4878,[48]2.4714,[49]2.4609,[50]2.4596,[51]2.4517,[52]2.4627,[53]2.4595,[54]2.4573,[55]2.4420,[56]2.4292,[57]2.4210,[58]2.4185,[59]2.4149,[60]2.4505,[61]2.4709,[62]2.4773,[63]2.4710,[64]2.4680,[65]2.4773,[66]2.4720,[67]2.4720,[68]2.4673,[69]2.4619,[70]2.4705,[71]2.5036,[72]2.5078,[73]2.5154,[74]2.5032,[75]2.5011,[76]2.5143,[77]2.4925,[78]2.4888,[79]2.4935,[80]2.4768,[81]2.4898,[82]2.4841,[83]2.4772,[84]2.4672,[85]2.4750,[86]2.4650,[87]2.4597,[88]2.4542,[89]2.4677,[90]2.4572,[91]2.4597,[92]2.4693,[93]2.4624,[94]2.4560,[95]2.4545,[96]2.4391,[97]2.4297,[98]2.4352,[99]2.4287,[100]2.4350,[101]2.4432,[102]2.4441,[103]2.4464,[104]2.4414,[105]2.4460,[106]2.4576,[107]2.4758,[108]2.4782,[109]2.4805,[110]2.4825,[111]2.4714,[112]2.4758,[113]2.4681,[114]2.4712,[115]2.4603,[116]2.4665,[117]2.4638,[118]2.4583,[119]2.4746,[120]2.4680,[121]2.5093,[122]2.5251,[123]2.5421,[124]2.5399,[125]2.5322,[126]2.5254,[127]2.5155,[128]2.5236,[129]2.5435,[130]2.5428,[131]2.5536,[132]2.5448,[133]2.5420,[134]2.5412,[135]2.5546,[136]2.5637,[137]2.5818,[138]2.6045,[139]2.6121,[140]2.6175,[141]2.6103,[142]2.6103,[143]2.6125,[144]2.6101,[145]2.6147,[146]2.6344,[147]2.6396,[148]2.6424,[149]2.6362,[150]2.6296,[151]2.6259,[152]2.6199,[153]2.6135,[154]2.6215,[155]2.6232,[156]2.6297,[157]2.6228,[158]2.6216,[159]2.6179,[160]2.6257,[161]2.6377,[162]2.6323,[163]2.6354,[164]2.6350,[165]2.6284,[166]2.6253,[167]2.6244,[168]2.6232,[169]2.6099,[170]2.6120,[171]2.6085,[172]2.6026,[173]2.5962,[174]2.5899,[175]2.5845,[176]2.5845,[177]2.5799,[178]2.5863,[179]2.5881,[180]2.5817,[181]2.5843,[182]2.5858,[183]2.5946,[184]2.5939,[185]2.5866,[186]2.5797,[187]2.5783,[188]2.5764,[189]2.5719,[190]2.5679,[191]2.5628,[192]2.5662,[193]2.5668,[194]2.5668,[195]2.5678,[196]2.5746,[197]2.5740,[198]2.5685,[199]2.5642,[200]2.5646,[201]2.5596,[202]2.5638,[203]2.5660,[204]2.5659,[205]2.5666,[206]2.5753,[207]2.5715,[208]2.5640,[209]2.5581,[210]2.5524,
|
| 3 |
+
|
eval/ppl-aria-gemma4-31b-Q5_K_M.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
2.03 minutes
|
| 2 |
+
[1]2.2266,[2]4.5475,[3]3.9285,[4]3.4593,[5]3.2633,[6]3.3352,[7]3.2738,[8]3.2358,[9]3.0840,[10]3.0809,[11]2.9228,[12]2.7663,[13]2.7077,[14]2.5905,[15]2.6286,[16]2.6257,[17]2.5603,[18]2.5136,[19]2.4249,[20]2.3866,[21]2.3774,[22]2.3553,[23]2.3446,[24]2.3349,[25]2.3246,[26]2.3429,[27]2.3475,[28]2.3075,[29]2.3091,[30]2.3031,[31]2.3171,[32]2.3184,[33]2.3190,[34]2.2989,[35]2.3228,[36]2.3759,[37]2.3670,[38]2.3582,[39]2.3472,[40]2.3318,[41]2.3060,[42]2.3142,[43]2.3734,[44]2.4404,[45]2.4397,[46]2.4357,[47]2.4469,[48]2.4312,[49]2.4210,[50]2.4208,[51]2.4139,[52]2.4251,[53]2.4227,[54]2.4211,[55]2.4063,[56]2.3939,[57]2.3863,[58]2.3823,[59]2.3789,[60]2.4147,[61]2.4352,[62]2.4412,[63]2.4345,[64]2.4319,[65]2.4408,[66]2.4356,[67]2.4354,[68]2.4307,[69]2.4253,[70]2.4336,[71]2.4663,[72]2.4708,[73]2.4781,[74]2.4655,[75]2.4637,[76]2.4766,[77]2.4554,[78]2.4517,[79]2.4554,[80]2.4398,[81]2.4528,[82]2.4471,[83]2.4406,[84]2.4306,[85]2.4385,[86]2.4284,[87]2.4231,[88]2.4180,[89]2.4313,[90]2.4211,[91]2.4234,[92]2.4325,[93]2.4255,[94]2.4194,[95]2.4183,[96]2.4027,[97]2.3933,[98]2.3991,[99]2.3926,[100]2.3994,[101]2.4077,[102]2.4078,[103]2.4107,[104]2.4061,[105]2.4107,[106]2.4225,[107]2.4400,[108]2.4422,[109]2.4447,[110]2.4470,[111]2.4362,[112]2.4407,[113]2.4339,[114]2.4366,[115]2.4254,[116]2.4316,[117]2.4291,[118]2.4234,[119]2.4385,[120]2.4318,[121]2.4726,[122]2.4874,[123]2.5040,[124]2.5020,[125]2.4945,[126]2.4874,[127]2.4778,[128]2.4855,[129]2.5052,[130]2.5047,[131]2.5153,[132]2.5062,[133]2.5035,[134]2.5024,[135]2.5153,[136]2.5237,[137]2.5418,[138]2.5638,[139]2.5710,[140]2.5762,[141]2.5689,[142]2.5698,[143]2.5727,[144]2.5706,[145]2.5749,[146]2.5945,[147]2.5995,[148]2.6026,[149]2.5965,[150]2.5899,[151]2.5858,[152]2.5793,[153]2.5728,[154]2.5809,[155]2.5826,[156]2.5889,[157]2.5824,[158]2.5814,[159]2.5781,[160]2.5857,[161]2.5972,[162]2.5919,[163]2.5947,[164]2.5944,[165]2.5878,[166]2.5846,[167]2.5832,[168]2.5822,[169]2.5690,[170]2.5711,[171]2.5674,[172]2.5617,[173]2.5557,[174]2.5495,[175]2.5443,[176]2.5446,[177]2.5403,[178]2.5464,[179]2.5481,[180]2.5421,[181]2.5444,[182]2.5457,[183]2.5542,[184]2.5539,[185]2.5466,[186]2.5399,[187]2.5386,[188]2.5365,[189]2.5319,[190]2.5281,[191]2.5230,[192]2.5263,[193]2.5269,[194]2.5268,[195]2.5276,[196]2.5342,[197]2.5331,[198]2.5279,[199]2.5238,[200]2.5238,[201]2.5192,[202]2.5231,[203]2.5250,[204]2.5246,[205]2.5254,[206]2.5338,[207]2.5300,[208]2.5226,[209]2.5171,[210]2.5117,
|
| 3 |
+
|
eval/ppl-aria-gemma4-31b-Q8_0.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
12.07 minutes
|
| 2 |
+
[1]2.2200,[2]4.5315,[3]3.9305,[4]3.4430,[5]3.2373,[6]3.3103,[7]3.2569,[8]3.2219,[9]3.0708,[10]3.0655,[11]2.9098,[12]2.7565,[13]2.7008,[14]2.5842,[15]2.6236,[16]2.6209,[17]2.5553,[18]2.5077,[19]2.4202,[20]2.3798,[21]2.3696,[22]2.3468,[23]2.3342,[24]2.3240,[25]2.3138,[26]2.3326,[27]2.3374,[28]2.2976,[29]2.2986,[30]2.2925,[31]2.3058,[32]2.3085,[33]2.3086,[34]2.2894,[35]2.3136,[36]2.3677,[37]2.3588,[38]2.3490,[39]2.3385,[40]2.3233,[41]2.2976,[42]2.3049,[43]2.3637,[44]2.4310,[45]2.4304,[46]2.4265,[47]2.4376,[48]2.4218,[49]2.4111,[50]2.4106,[51]2.4034,[52]2.4144,[53]2.4116,[54]2.4095,[55]2.3947,[56]2.3825,[57]2.3749,[58]2.3713,[59]2.3679,[60]2.4032,[61]2.4235,[62]2.4290,[63]2.4224,[64]2.4198,[65]2.4291,[66]2.4240,[67]2.4236,[68]2.4190,[69]2.4137,[70]2.4218,[71]2.4544,[72]2.4591,[73]2.4662,[74]2.4537,[75]2.4517,[76]2.4648,[77]2.4436,[78]2.4399,[79]2.4439,[80]2.4288,[81]2.4418,[82]2.4362,[83]2.4297,[84]2.4196,[85]2.4278,[86]2.4179,[87]2.4126,[88]2.4075,[89]2.4212,[90]2.4107,[91]2.4131,[92]2.4224,[93]2.4155,[94]2.4090,[95]2.4081,[96]2.3924,[97]2.3831,[98]2.3889,[99]2.3823,[100]2.3890,[101]2.3972,[102]2.3973,[103]2.3997,[104]2.3950,[105]2.3995,[106]2.4113,[107]2.4292,[108]2.4315,[109]2.4339,[110]2.4363,[111]2.4256,[112]2.4299,[113]2.4227,[114]2.4257,[115]2.4146,[116]2.4206,[117]2.4181,[118]2.4124,[119]2.4275,[120]2.4211,[121]2.4620,[122]2.4772,[123]2.4937,[124]2.4911,[125]2.4833,[126]2.4763,[127]2.4666,[128]2.4745,[129]2.4943,[130]2.4937,[131]2.5043,[132]2.4952,[133]2.4926,[134]2.4918,[135]2.5048,[136]2.5133,[137]2.5316,[138]2.5534,[139]2.5607,[140]2.5663,[141]2.5590,[142]2.5597,[143]2.5623,[144]2.5599,[145]2.5644,[146]2.5839,[147]2.5890,[148]2.5922,[149]2.5859,[150]2.5793,[151]2.5752,[152]2.5689,[153]2.5625,[154]2.5706,[155]2.5723,[156]2.5788,[157]2.5719,[158]2.5708,[159]2.5674,[160]2.5753,[161]2.5869,[162]2.5816,[163]2.5844,[164]2.5840,[165]2.5775,[166]2.5742,[167]2.5729,[168]2.5719,[169]2.5589,[170]2.5610,[171]2.5572,[172]2.5515,[173]2.5454,[174]2.5394,[175]2.5341,[176]2.5343,[177]2.5300,[178]2.5360,[179]2.5380,[180]2.5319,[181]2.5342,[182]2.5355,[183]2.5442,[184]2.5438,[185]2.5366,[186]2.5300,[187]2.5287,[188]2.5267,[189]2.5220,[190]2.5183,[191]2.5130,[192]2.5165,[193]2.5169,[194]2.5170,[195]2.5179,[196]2.5245,[197]2.5237,[198]2.5184,[199]2.5142,[200]2.5144,[201]2.5097,[202]2.5137,[203]2.5157,[204]2.5153,[205]2.5160,[206]2.5246,[207]2.5207,[208]2.5135,[209]2.5078,[210]2.5023,
|
| 3 |
+
|
eval/text_metadata.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_dir": "/alloc/gemma4_full_sft_recipe/runs/aria_gemma4_31b_identity_rerun_lr2e6_ep5_ga4_anchor10_20260414_094544/merged_epoch2_best",
|
| 3 |
+
"seed": 42,
|
| 4 |
+
"calibration_samples": 256,
|
| 5 |
+
"eval_samples": 200,
|
| 6 |
+
"calibration_identity_samples": 192,
|
| 7 |
+
"calibration_anchor_samples": 64,
|
| 8 |
+
"eval_identity_samples": 100,
|
| 9 |
+
"eval_wildchat_samples": 100,
|
| 10 |
+
"eval_from_train": true
|
| 11 |
+
}
|