Downloads · 30 days
142
7% of all-time downloads
VibeThinker-3B-Function-Calling-xLAM-Unsloth-GGUF
ermiaazarkhalili/VibeThinker-3B-Function-Calling-xLAM-Unsloth-GGUF
VibeThinker-3B-Function-Calling-xLAM-Unsloth-GGUF is a text generation model from ermiaazarkhalili. Use it when you need the model to write or continue text. It is set up for gguf. The card lists the license as mit.
GGUF quantizations of a LoRA fine-tune of WeiboAI/VibeThinker-3B, supervised fine-tuned on Salesforce/xlam-function-calling-60k.
All-time downloads
2.1K
Public
Repo size
12.8 GB
Likes
1
Public
Repo makeup
Click a slice to open those files.
.gguf12.8 GB · 100%
At a glance
- Task
- Text Generation
- Library
- gguf
- License
- mit
- Access
- Public
- Created
- Jun 22, 2026
- Updated
- Aug 23, 2026
- SHA
- 62bfdf30
Try a prompt
Base models
Model card
From the Hugging Face model README
VibeThinker-3B-Function-Calling-xLAM-Unsloth-GGUF
GGUF quantizations of a LoRA fine-tune of WeiboAI/VibeThinker-3B, supervised fine-tuned on Salesforce/xlam-function-calling-60k.
Quantized from ermiaazarkhalili/VibeThinker-3B-Function-Calling-xLAM-Unsloth. See that repository for the full-precision weights.
| Base model | WeiboAI/VibeThinker-3B |
| Training data | Salesforce/xlam-function-calling-60k |
| Method | LoRA supervised fine-tuning via Unsloth + TRL |
| License | mit (inherited from the base model) |
Available quantizations
| File | Size |
|---|---|
vibethinker-3b-function-calling-xlam-unsloth.q2_k.gguf | 1.27 GB |
vibethinker-3b-function-calling-xlam-unsloth.q3_k_m.gguf | 1.59 GB |
vibethinker-3b-function-calling-xlam-unsloth.q4_k_m.gguf | 1.93 GB |
vibethinker-3b-function-calling-xlam-unsloth.q5_k_m.gguf | 2.22 GB |
vibethinker-3b-function-calling-xlam-unsloth.q6_k.gguf | 2.54 GB |
vibethinker-3b-function-calling-xlam-unsloth.q8_0.gguf | 3.29 GB |
Usage
llama.cpp
huggingface-cli download ermiaazarkhalili/VibeThinker-3B-Function-Calling-xLAM-Unsloth-GGUF vibethinker-3b-function-calling-xlam-unsloth.q4_k_m.gguf --local-dir .
llama-cli -m vibethinker-3b-function-calling-xlam-unsloth.q4_k_m.gguf -p "Explain gradient checkpointing in two sentences." -n 256
Ollama
echo 'FROM ./vibethinker-3b-function-calling-xlam-unsloth.q4_k_m.gguf' > Modelfile
ollama create vibethinker-3b-function-calling-xlam-unsloth-gguf -f Modelfile
ollama run vibethinker-3b-function-calling-xlam-unsloth-gguf
Training configuration
| Setting | Value |
|---|---|
| LoRA rank (r) | 16 |
| LoRA alpha | 16 |
| Learning rate | 0.0002 |
| Epochs | 1 |
| Effective batch size | 8 (2 x 4 grad accum) |
| Max sequence length | 2048 |
| Base precision | 4-bit (QLoRA) |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
Observed training loss
Measured from our SLURM logs for this configuration. These are training-loss observations only — no downstream benchmark evaluation has been run on this model, so they should not be read as a quality claim.
| SLURM job | Steps | First loss | Final loss |
|---|---|---|---|
45169146 | 7,500 | 1.0532 | 0.1784 |
Limitations
- No benchmark evaluation has been run on this checkpoint. The only reported numbers are training-loss observations.
- Inherits the biases, knowledge cutoff and failure modes of the base model.
- Fine-tuned on a single instruction-following dataset; behaviour outside that distribution is untested.
- LoRA adapters were merged into the base weights, so the merged model cannot be detached from this fine-tune.
Reproducing
Trained by notebooks/xlam_function_calling_vibethinker-3b_unsloth.ipynb, executed non-interactively with
papermill on a SLURM H100 partition (Unsloth + TRL, LoRA).
Card generated from the training run's own configuration and logs by
scripts/generate_hub_model_card.py.
- Task
- Text Generation
- Library
- gguf
- License
- mit
- Created
- Jun 22, 2026
- Updated
- Aug 23, 2026