ibm-granite/granite-20b-code-instruct-8k
Full fine-tune vs ibm-granite/granite-20b-code-base-8k. Fine-tune: 100% of the weights changed, across 52 of 52 layers.
Tensors changed: Attention 208/208, Norms 210/210, MLP 208/208, Embeddings 2/2
7 derivatives checked
- bartowski/granite-20b-code-instruct-GGUF: Same weights, converted
- mradermacher/granite-20b-code-instruct-i1-GGUF: Full fine-tune
- mradermacher/granite-20b-code-instruct-8k-i1-GGUF: Quantization
- legraphista/granite-20b-code-instruct-IMat-GGUF: Same weights, converted
- mradermacher/granite-20b-code-instruct-8k-GGUF: Quantization
- mradermacher/granite-20b-code-instruct-GGUF: Full fine-tune
- DevQuasar-2/granite-20b-code-instruct-8k-GGUF: Quantization
2 other repos hold byte-for-byte identical weight files.