Downloads · 30 days
332
12% of all-time downloads
ermiaazarkhalili/Qwen3.8-9B-SFT-Fable5-Glint-GGUF
Qwen3.8-9B-SFT-Fable5-Glint-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 apache-2.0.
GGUF quantizations of a LoRA fine-tune of empero-ai/Qwen3.8-9B, supervised fine-tuned on ermiaazarkhalili/Fable-5-Glint-Clean (private).
Downloads · 30 days
332
12% of all-time downloads
All-time downloads
2.8K
Public
Repo size
38.4 GB
Likes
3
Public
Click a slice to open those files.
.gguf38.4 GB · 100%
From the Hugging Face model README
GGUF quantizations of a LoRA fine-tune of empero-ai/Qwen3.8-9B, supervised fine-tuned on ermiaazarkhalili/Fable-5-Glint-Clean (private).
Quantized from ermiaazarkhalili/Qwen3.8-9B-SFT-Fable5-Glint. See that repository for the full-precision weights.
| Base model | empero-ai/Qwen3.8-9B |
| Training data | ermiaazarkhalili/Fable-5-Glint-Clean (private) |
| Method | LoRA supervised fine-tuning via Unsloth + TRL |
| License | apache-2.0 (inherited from the base model) |
| File | Size |
|---|---|
qwen3.8-9b-sft-fable5-glint.q2_k.gguf | 3.91 GB |
qwen3.8-9b-sft-fable5-glint.q3_k_m.gguf | 4.74 GB |
qwen3.8-9b-sft-fable5-glint.q4_k_m.gguf | 5.78 GB |
qwen3.8-9b-sft-fable5-glint.q5_k_m.gguf | 6.64 GB |
qwen3.8-9b-sft-fable5-glint.q6_k.gguf | 7.56 GB |
qwen3.8-9b-sft-fable5-glint.q8_0.gguf | 9.79 GB |
huggingface-cli download ermiaazarkhalili/Qwen3.8-9B-SFT-Fable5-Glint-GGUF qwen3.8-9b-sft-fable5-glint.q4_k_m.gguf --local-dir .
llama-cli -m qwen3.8-9b-sft-fable5-glint.q4_k_m.gguf -p "Explain gradient checkpointing in two sentences." -n 256
echo 'FROM ./qwen3.8-9b-sft-fable5-glint.q4_k_m.gguf' > Modelfile
ollama create qwen3.8-9b-sft-fable5-glint-gguf -f Modelfile
ollama run qwen3.8-9b-sft-fable5-glint-gguf
| Setting | Value |
|---|---|
| LoRA rank (r) | 16 |
| LoRA alpha | 16 |
| Learning rate | 0.0002 |
| Epochs | 3 |
| Effective batch size | 8 (1 x 8 grad accum) |
| Max sequence length | 4096 |
| Base precision | 4-bit (QLoRA) |
| Target modules | down_proj, gate_proj, in_proj_a, in_proj_b, in_proj_qkv, in_proj_z, k_proj, o_proj, out_proj, q_proj, up_proj, v_proj |
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 |
|---|---|---|---|
55541066 | 1,554 | 0.9887 | 0.6265 |
Trained by notebooks/fable_distillation_qwen38-9b_fable-glint_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.