finetune-lora: Add checkpoint saving & resuming from saved checkpoint #32
+720
−58
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit adds checkpointing for fine-tuning: - Add checkpoint saving every N steps with --checkpoint-save-steps - Save complete training state: model weights, optimizer state, metadata - Implement two-phase optimizer state loading to avoid memory issues - Add --resume-from and --auto-resume functionality - Store optimizer momentum/variance tensors in GGUF format - Add checkpoint validation for rank, alpha, and target modules - Update README.md with checkpointing documentation The optimizer state loading: iteration count is loaded during initialization, while tensor data (grad_m, grad_v) is loaded after ggml_opt_alloc creates the proper tensor structures.
Signed-off-by: Marcus Edel <[email protected]>
Signed-off-by: Marcus Edel <[email protected]>
…sion using .string().c_str(). Signed-off-by: Marcus Edel <[email protected]>
olyasir
approved these changes
Oct 16, 2025
gianni-cor
reviewed
Oct 16, 2025
| params.escape = false; | ||
| parse_finetune_args(argc, argv, ft_params); | ||
|
|
||
| if (!common_params_parse(argc, argv, params, LLAMA_EXAMPLE_PERPLEXITY)) { |
There was a problem hiding this comment.
can we make all parameters from LLAMA_EXAMPLE_PERPLEXITY also be available in LLAMA_EXAMPLE_MAIN ? as we use this last one in the addon to access all tunable parameters.
makaveli10
changed the base branch from
temp-latest
to
temp-latest-finetuning
olyasir
reviewed
Oct 17, 2025
| // get the gradient accumulator for a node from the forward graph | ||
| GGML_API struct ggml_tensor * ggml_opt_grad_acc(ggml_opt_context_t opt_ctx, struct ggml_tensor * node); | ||
|
|
||
| // get optimizer state tensors (momentum and variance for AdamW) |
There was a problem hiding this comment.
Can we also add SGD support, so we could choose between them on command line? SGD have fewer parameters, and looks like it is already supported in llama.cpp
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR adds checkpointing for fine-tuning:
The optimizer state loading: iteration count is loaded during initialization, while tensor data (grad_m, grad_v) is loaded after ggml_opt_alloc creates the proper tensor structures.