ibm-granite/granite-3.3-8b-instruct
Full fine-tune vs ibm-granite/granite-3.3-8b-base. Fine-tune: 100% of the weights changed, across 40 of 40 layers.
Tensors changed: Norms 81/81, MLP 120/120, Attention 160/160
19 derivatives checked
- ibm-ai-platform/micro-g3.3-8b-instruct-1b: Full fine-tune (fewer layers)
- ibm-granite/granite-speech-3.3-8b: Same weights, repacked (fewer layers)
- ibm-granite/granite-3.3-8b-instruct-GGUF: Same weights, converted
- ibm-granite/granite-3.3-8b-instruct-FP8: Quantization
- bartowski/ibm-granite_granite-3.3-8b-instruct-GGUF: Same weights, converted
- unsloth/granite-3.3-8b-instruct-GGUF: Same weights, converted
- mradermacher/granite-3.3-8b-instruct-i1-GGUF: Quantization
- ibm-granite/granite-3.3-8b-math-prm-v2: Full fine-tune
- Raymond-dev-546730/ClinicalThought-AI-8B: Full fine-tune
- DevQuasar-12/ibm-granite.granite-3.3-8b-instruct-GGUF: Same weights, converted
- mradermacher/granite-3.3-8b-instruct-GGUF: Same weights, converted
- mlx-community/granite-3.3-8b-instruct-4bit: Quantization
- lmstudio-community/granite-3.3-8b-instruct-GGUF: Quantization
- lucas-sousa-pereira/granite-3-3-8b-java-code-refactoring: Full fine-tune
- drawais/Granite-3.3-8B-Instruct-AWQ-INT4: Quantization
- mlx-community/granite-3.3-8b-instruct-fp16: Same weights, converted
- Qiskit/granite-3.3-8b-qiskit: Full fine-tune
- heterodoxin/granite-3.3-8b-instruct-apostate: Abliterated
- alexj03/granite-3.3-8b-instruct-abliterated: Partial fine-tune
3 other repos hold byte-for-byte identical weight files.