mirror of
https://github.com/ggerganov/llama.cpp
synced 2026-03-15 19:51:26 +01:00
e8d99dd0b6
* feat: Add NEMOTRONH to python arch enum https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add NEMOTRONH to c++ arch enum https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add NEMOTRONH to llama-arch layer map https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: First pass at conversion for nemotronh https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: Add a verbose log for each tensor loaded This is really helpful for diagnosing mismatches between the expected and received tensors https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * feat: First (broken) pass at nemotronh model architecture It generates tokens, just not valid ones! https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Explicitly enable add_bos_token during conversion The `tokenizer.json`/`tokenizer_config.json` in the model are a bit contradictory. In the config, add_bos_token is set to False, but the tokenizer model itself has a post_processor that adds the BOS token via type: TemplateProcessing https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use relu2 (LLM_FFN_RELU_SQR) for activation in FFN layers https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Only allocate attention cache for attention layers (not non-recurrent) https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Move residual add to after every block https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * fix: Use the correct norm tensor for the MLP blocks https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> * Nemotron-H: MLP gate cleanup (pass NULL for unused gate) This model does not use a gate in MLP blocks; pass NULLs for gate tensors to make intent clear and avoid unused-pointer noise. * SSM: respect ssm_dt_rank for dt_dim when provided Use GGUF-provided time_step_rank (ssm_dt_rank) to set dt_dim when > 0; fallback to max(64, n_embd/16). * fix: plamo2 - revert dt_dim to default (remove ssm_dt_rank usage) * Rename nemotronh to nemotron_h for consistency - Update architecture name from NEMOTRONH to NEMOTRON_H in constants.py - Change architecture string from 'nemotronh' to 'nemotron_h' in all files - Update enum LLM_ARCH_NEMOTRONH to LLM_ARCH_NEMOTRON_H - Update class name llm_build_nemotronh to llm_build_nemotron_h - Consistent naming with underscore convention (nemotron_h vs nemotronh) * feat: Support conversion for older NemotronH models https://github.com/ggml-org/llama.cpp/issues/nemotron-nano-15409 Branch: gabe-l-hart/nvidia-nemotron-nano-15409 Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> --------- Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> Co-authored-by: Maicon Domingues <dominguesm@outlook.com> Co-authored-by: weatherman <fxdstudios@gmail.com> |
||
|---|---|---|
| .. | ||
| CMakeLists.txt | ||
| llama-adapter.cpp | ||
| llama-adapter.h | ||
| llama-arch.cpp | ||
| llama-arch.h | ||
| llama-batch.cpp | ||
| llama-batch.h | ||
| llama-chat.cpp | ||
| llama-chat.h | ||
| llama-context.cpp | ||
| llama-context.h | ||
| llama-cparams.cpp | ||
| llama-cparams.h | ||
| llama-grammar.cpp | ||
| llama-grammar.h | ||
| llama-graph.cpp | ||
| llama-graph.h | ||
| llama-hparams.cpp | ||
| llama-hparams.h | ||
| llama-impl.cpp | ||
| llama-impl.h | ||
| llama-io.cpp | ||
| llama-io.h | ||
| llama-kv-cache-iswa.cpp | ||
| llama-kv-cache-iswa.h | ||
| llama-kv-cache.cpp | ||
| llama-kv-cache.h | ||
| llama-kv-cells.h | ||
| llama-memory-hybrid.cpp | ||
| llama-memory-hybrid.h | ||
| llama-memory-recurrent.cpp | ||
| llama-memory-recurrent.h | ||
| llama-memory.cpp | ||
| llama-memory.h | ||
| llama-mmap.cpp | ||
| llama-mmap.h | ||
| llama-model-loader.cpp | ||
| llama-model-loader.h | ||
| llama-model-saver.cpp | ||
| llama-model-saver.h | ||
| llama-model.cpp | ||
| llama-model.h | ||
| llama-quant.cpp | ||
| llama-quant.h | ||
| llama-sampling.cpp | ||
| llama-sampling.h | ||
| llama-vocab.cpp | ||
| llama-vocab.h | ||
| llama.cpp | ||
| unicode-data.cpp | ||
| unicode-data.h | ||
| unicode.cpp | ||
| unicode.h | ||