Nvflare Convert Huggingface

NVFLARE Convert Hugging Face

Use When

Use when converting training code built around transformers.Trainer, Seq2SeqTrainer, TRL SFTTrainer, or another Trainer subclass. Support full-model and PEFT/LoRA fine-tuning, datasets/tokenizers, Trainer callbacks and metrics, checkpoint continuity, and replicated torch.distributed training.

Do Not Use When

Do not use for an AutoModel driven by a manual PyTorch loop without a Hugging Face Trainer (route to nvflare-convert-pytorch), PyTorch Lightning (route to nvflare-convert-lightning, including Lightning modules that contain Transformers models), inference-only pipelines, model serving, failed jobs (route to nvflare-diagnose-job), or federated statistics without training (route to nvflare-fed-stats). Route a project with active Lightning and Hugging Face Trainer entrypoints to nvflare-orient to select one training-loop owner or separate jobs. Route unresolved Trainer ownership, such as a Trainer factory without a bound owner call, to nvflare-orient; do not patch either Trainer. Out of scope: DeepSpeed, FSDP, production/POC deployment, controller rewrites, experiment search, and privacy-protection requests such as HE, encrypted aggregation, differential privacy, or privacy filters; never substitute an unprotected recipe or present a disclaimer as implementation. If a request combines federated statistics and model-training conversion, treat it as two independent jobs and workflows: do not merge or automatically chain them, do not route the combination to nvflare-orient, and ask which workflow to run first before generating or running either job. Recommend nvflare-fed-stats first only when the user's purpose is to understand data distribution; handle conversion later as a separate request.

Available Scripts

Use run_script() only during the validation phase after loading references/huggingface-validation.md. For example:

• run_script("scripts/resolve_model_snapshot.py", ["--source", "local", "--source-root", "<absolute-source-root>", "<configured-path>"])

• run_script("scripts/resolve_model_snapshot.py", ["--source", "hub", "<org/model>"])

Workflow

1. Load ../nvflare-shared/references/conversion-common.md and apply it for the whole conversion; this SKILL.md states only the framework-specific deltas. Load ../nvflare-shared/references/conversion-workflow.md only for a non-standard rerun, authorization, or missing-semantics case; it no longer holds the data-location or partitioning contracts, whose invariants conversion-common.md owns. Load ../nvflare-shared/references/site-data-and-paths.md for generated partitions, relative paths, or per-site data locations.

2. Inspect before editing with nvflare agent inspect source <path> --format json plus direct source reading. Load references/huggingface-detection.md during this phase. If inspect recommends nvflare-orient for unresolved Trainer ownership or active Lightning/Hugging Face owners, stop before editing. Extract the entrypoint, Trainer subclass, model constructor, tokenizer or processor, datasets and collator, Trainer arguments, compute_metrics, callbacks, checkpoint and PEFT settings, precision, local budget, distributed launcher, site/round counts, data location, and aggregation intent. Do not import or execute user training modules to discover them.

3. Apply the dependency-install ordering rule in ../nvflare-shared/references/conversion-common.md before any Python command imports user, framework, NVFLARE, or declared dependencies. Keep dependency inventory single-purpose: check NVFLARE separately, inventory non-product packages separately, and do not append Hugging Face cache or filesystem discovery. Defer model and dataset availability checks to the maintained resolver during validation. If an optional path must be inspected, run it separately and report a missing directory with exit code zero.

4. Select the recipe from FL intent. For explicit FedAvg, run nvflare recipe show fedavg-pt --format json, then immediately load ../nvflare-shared/references/pytorch-family-recipe-construction.md and use the returned module, class, and parameters with the required construction and execution shape in assets/job.py. Import FedAvgRecipe from nvflare.app_opt.pt.recipes.fedavg, never from nvflare.recipe. Treat class_path as the public recipe key and path as its normalized exported representation; do not inspect Recipe source or signatures to reconcile them. Do not guess adjacent symbols or add per-site recipe config unless sites genuinely differ. Load ../nvflare-shared/references/pytorch-family-recipe-selection.md only for ambiguous, evaluation-only, or non-FedAvg requests.

5. Convert with references/huggingface-conversion.md and adapt assets/client_with_eval.py rather than drafting a new round loop. Preserve model, tokenizer/processor, datasets, collator, Trainer arguments, callbacks, and metrics. Apply the step-1 data-location rules to the client's data argument. Import the Client API as import nvflare.client.hf as flare, so flare.init(), flare.patch(), and flare.is_running() resolve to nvflare.client.hf. Keep flare.patch(trainer) simple with inferred params_scope="auto" and encode one per-round budget in Trainer arguments: requested steps use max_steps, requested epochs use num_train_epochs, and a silent prompt uses the reported default max_steps=10 unless source-budget preservation was requested. Do not duplicate the budget in patch local_steps/local_epochs. When the client uses HfArgumentParser, construct it with allow_abbrev=False.

6. Adapt assets/server_model.py and assets/job.py instead of inventing server-model, packaging, export, or SimEnv wiring. Keep generated and packaged project-local modules in the same writable source directory. Never use .. in train_script, add_server_file(), or add_client_file(); use an existing resolved absolute path when co-location is impossible. Keep the server and Trainer model factory and exchange keyspace identical. Use the recipe's documented class_path or path key plus complete args for required or overridden values; a direct zero-argument instance must not use from_pretrained(), downloads, or checkpoint loading during job construction. Apply only options confirmed by the construction reference. Preserve the job asset's recipe-before-parser ordering, ArgumentParser(allow_abbrev=False), and strict parse_args(); do not use parse_known_args().

7. Only after generated files exist, load ../nvflare-shared/references/validation-evidence.md and references/huggingface-validation.md. Follow the shared compile, construction, simulation, and terminal-evidence ladder. Inspect export/package evidence only for an exported final target; inspect a local target's materialized evidence after its run. Apply only the standard HF Trainer checks and stop at the first failed rung. Review and exercise the maintained assets directly; do not inspect NVFLARE implementation source, improvise Recipe API probes, or write one-off AST programs to re-prove them. Use references/huggingface-state-and-distributed.md only when inspection found PEFT, DDP, checkpoint/restore overrides, auxiliary trainable models, or another non-default patch setting.

8. Report the recipe, source facts, parameter scope, data partition, changed files, validation status, and exact artifact paths. When validation produces metrics, load ../nvflare-shared/references/metrics-and-artifact-reporting.md before the final response and report each observed primary scalar with its metric name, numeric value, and artifact or bounded-log source.

Requirements

• Must use flare.patch(trainer) as the sole model-exchange owner. receive() inside a patched loop may inspect task metadata only; it must not load a second copy of the global model.

• Must follow the Client API initialization and conditional rank contract in ../nvflare-shared/references/conversion-common.md. Keep the generated client rankless; nvflare.client.hf.init() owns distributed-rank resolution and rejection. Load references/huggingface-state-and-distributed.md only for a distributed multi-process source path.

• Must preserve source evaluation. When per-round global-model evaluation is required, call trainer.evaluate() before trainer.train() on every rank. Do not invent compute_metrics, label mappings, averaging denominators, or metric direction.

• Must follow the Best-Model Metric policy in ../nvflare-shared/references/pytorch-family-recipe-construction.md; the Hugging Face delta is only how the delivered key is named and produced. Must preserve source metric names when practical: if the generated trainer.evaluate() emits accuracy, set key_metric="accuracy"; if Trainer emits a prefixed key such as eval_accuracy, set the server to that exact key and report the source-to-server mapping. When best-model selection is requested, every lower-is-better metric, including Trainer-generated eval_loss, is delivered as an explicitly negated companion and selected by that key — never as raw loss. When selection is not requested, use key_metric=""; do not omit it and accidentally activate the recipe default.

• Must preserve PEFT configuration exactly and verify adapter key compatibility between the server model and patched Trainer. Do not infer LoRA target modules, silently switch adapter/full-model scope, or solve key mismatches with non-strict loading.

• Must verify that trainer.model owns all federated trainable state for Trainer subclasses with reference, reward, value-head, or other auxiliary models. Ask or fail closed when params_scope="auto" would omit trainable state required by the algorithm.

• Must preserve model constructor values needed on both server and clients per ../nvflare-shared/references/pytorch-model-exchange.md (State-Dict Compatibility). Ask one semantic question or fail closed when required values are not statically available.

• Must patch only one Trainer per Python process. Preserve a single Trainer lifecycle across rounds when restore_state=True.

• Must use a positive TrainingArguments.max_steps budget for a length-less iterable training dataset and let flare.patch(trainer) infer it.

• Must reject or report DeepSpeed, FSDP, save_only_model=True with restore_state=True, load_best_model_at_end=True, prebuilt optimizer/scheduler instances with restore_state=False, and checkpoint paths not visible to every distributed rank. Do not rewrite these settings silently. launch_once is a framework-neutral recipe parameter owned by ../nvflare-shared/references/pytorch-family-recipe-construction.md; the Hugging Face delta is only that the product rejects explicit launch_once=False together with restore_state=True.

• Must initialize torch.distributed before patching when rank environment variables declare multiple ranks. All ranks must call patched Trainer methods in identical order.

• Must use the maintained HF validation resolver with an explicit local/Hub source. For authorized downloads it obtains or validates a full commit-SHA revision before downloading. Must not copy it into generated job code, set trust_remote_code=True, download model/data artifacts unless requested, or recover from a cache-only miss by going online. Cache misses, remote identifiers, and validation requests do not authorize online retries; see ../nvflare-shared/references/conversion-common.md.

• Site partitioning, custom aggregation, the Source Of Truth Boundary, and user input/authorization follow ../nvflare-shared/references/conversion-common.md.

Always read this converter SKILL.md together with ../nvflare-shared/references/conversion-common.md. Complete each workflow phase before loading the next phase's reference. Do not preload validation, state/DDP, broad workflow, dependency, or reporting references. The standard FedAvg path loads, in order: ../nvflare-shared/references/conversion-common.md, references/huggingface-detection.md, and ../nvflare-shared/references/site-data-and-paths.md only for its stated triggers; ../nvflare-shared/references/pytorch-family-recipe-construction.md, references/huggingface-conversion.md, and ../nvflare-shared/references/pytorch-model-exchange.md; then ../nvflare-shared/references/validation-evidence.md and references/huggingface-validation.md; load references/huggingface-state-and-distributed.md and other shared references only under the triggers above. Do not depend on repository examples.