id stringlengths 12 32 | archetype stringclasses 30
values | split stringclasses 1
value | prompt stringlengths 4.71k 5.31k | gold stringlengths 68 186 | meta stringclasses 30
values |
|---|---|---|---|---|---|
train-shard_size-041 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 25000, "shard_size_gb": 25.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-checkpoint_retention-056 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 6}, "work": {"checkpoint_size_gb": 112.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-training_time-027 | training_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "TRAINING_TIME_ESTIMATE", "fix": {"planned_time_hours": 0.7111111111111111}, "work": {"total_tokens": 307200000}} | {"archetype": "training_time", "computations": ["R16"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_v100_dataloader-034 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-training_time-056 | training_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "TRAINING_TIME_ESTIMATE", "fix": {"planned_time_hours": 1.7777777777777777}, "work": {"total_tokens": 204800000}} | {"archetype": "training_time", "computations": ["R16"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_sampler_accum-017 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 8, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-two_oom_clipping-016 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 7}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-two_sampler_accum-021 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 4, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-two_lora_vocab-003 | two_lora_vocab | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["ADAPTERS_FROZEN_ZERO_TRAINABLE", "EMBEDDING_VOCAB_MISMATCH"], "fix": {"adapter_requires_grad": true, "embedding_vocab_size": 50304}} | {"archetype": "two_lora_vocab", "computations": ["R8", "R11"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-accum_steps_lr-012 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 0.0002325, "grad_accum_steps": 8, "planned_total_steps": 750}, "work": {"audit_step": 187, "steps_per_epoch": 250}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-two_oom_clipping-035 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 26}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-ckpt_batch_accum-063 | ckpt_batch_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINTED_BATCH_AND_ACCUM", "fix": {"grad_accum_steps": 2, "per_device_batch_size": 35}, "work": {"activation_per_sample_gb": 2.0, "max_per_device_batch": 35}} | {"archetype": "ckpt_batch_accum", "computations": ["R20", "R1", "R3"], "depth": 4, "interaction": "checkpointing x accum", "tier": "interacting"} |
train-two_oom_clipping-033 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 26}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-two_lora_vocab-016 | two_lora_vocab | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["ADAPTERS_FROZEN_ZERO_TRAINABLE", "EMBEDDING_VOCAB_MISMATCH"], "fix": {"adapter_requires_grad": true, "embedding_vocab_size": 50304}} | {"archetype": "two_lora_vocab", "computations": ["R8", "R11"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-zero_batch-039 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 32}, "work": {"bytes_per_param_per_rank": 4.0, "fixed_memory_gb": 32.0, "remaining_memory_gb": 64.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-checkpoint_retention-015 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 15}, "work": {"checkpoint_size_gb": 48.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-accum_and_steps-045 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 1, "planned_total_steps": 750}, "work": {"steps_per_epoch": 250}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-effective_batch-000 | effective_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "EFFECTIVE_BATCH_MISMATCH", "fix": {"grad_accum_steps": 4}} | {"archetype": "effective_batch", "computations": ["R3"], "depth": 2, "interaction": null, "tier": "diagnosis"} |
train-shard_size-001 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 12500, "shard_size_gb": 12.5}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_oom_clipping-045 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 7}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-ckpt_batch_accum-018 | ckpt_batch_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINTED_BATCH_AND_ACCUM", "fix": {"grad_accum_steps": 2, "per_device_batch_size": 66}, "work": {"activation_per_sample_gb": 1.0, "max_per_device_batch": 66}} | {"archetype": "ckpt_batch_accum", "computations": ["R20", "R1", "R3"], "depth": 4, "interaction": "checkpointing x accum", "tier": "interacting"} |
train-checkpoint_retention-041 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 8}, "work": {"checkpoint_size_gb": 24.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-shard_size-060 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 25000, "shard_size_gb": 60.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-zero_batch-032 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 12}, "work": {"bytes_per_param_per_rank": 4.0, "fixed_memory_gb": 16.0, "remaining_memory_gb": 24.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-checkpoint_retention-063 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 20}, "work": {"checkpoint_size_gb": 48.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-effective_batch-002 | effective_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "EFFECTIVE_BATCH_MISMATCH", "fix": {"grad_accum_steps": 16}} | {"archetype": "effective_batch", "computations": ["R3"], "depth": 2, "interaction": null, "tier": "diagnosis"} |
train-sampler-007 | sampler | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "fix": {"sampler": "DistributedSampler"}} | {"archetype": "sampler", "computations": ["R7"], "depth": 1, "interaction": null, "tier": "diagnosis"} |
train-accum_steps_lr-041 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 0.00011, "grad_accum_steps": 8, "planned_total_steps": 800}, "work": {"audit_step": 400, "steps_per_epoch": 400}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-two_sampler_accum-055 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 4, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-optimizer_batch-031 | optimizer_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "OPTIMIZER_MEMORY_BATCH", "fix": {"per_device_batch_size": 60}, "work": {"bytes_per_param": 12, "fixed_memory_gb": 36.0}} | {"archetype": "optimizer_batch", "computations": ["R21", "R1"], "depth": 3, "interaction": "optimizer x memory", "tier": "interacting"} |
train-lr_at_step-034 | lr_at_step | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "LR_VALUE_AT_STEP", "fix": {"current_learning_rate": 0.00011}, "work": {"step_fraction": 0.5}} | {"archetype": "lr_at_step", "computations": ["R15"], "depth": 2, "interaction": null, "tier": "computed"} |
train-accum_and_steps-069 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 8, "planned_total_steps": 2000}, "work": {"steps_per_epoch": 500}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_sampler_accum-039 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 2, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-shard_size-021 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 25000, "shard_size_gb": 25.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_sampler_accum-053 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 8, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-two_v100_dataloader-012 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-checkpoint_retention-062 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 31}, "work": {"checkpoint_size_gb": 24.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-parallel_batch-048 | parallel_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PARALLEL_SHARDED_BATCH", "fix": {"per_device_batch_size": 85}, "work": {"fixed_memory_gb": 64.0, "params_per_rank_billions": 4.0}} | {"archetype": "parallel_batch", "computations": ["R22", "R13", "R1"], "depth": 4, "interaction": "parallelism x memory", "tier": "interacting"} |
train-shard_size-053 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 12500, "shard_size_gb": 25.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-zero_batch-067 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 24}, "work": {"bytes_per_param_per_rank": 8.0, "fixed_memory_gb": 32.0, "remaining_memory_gb": 48.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-two_sampler_accum-054 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 16, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-two_oom_clipping-051 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 28}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-lr_at_step-052 | lr_at_step | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "LR_VALUE_AT_STEP", "fix": {"current_learning_rate": 0.000155}, "work": {"step_fraction": 0.25}} | {"archetype": "lr_at_step", "computations": ["R15"], "depth": 2, "interaction": null, "tier": "computed"} |
train-oom_batch-000 | oom_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "OOM_BATCH_TOO_LARGE", "fix": {"per_device_batch_size": 5}} | {"archetype": "oom_batch", "computations": ["R1"], "depth": 2, "interaction": null, "tier": "diagnosis"} |
train-ckpt_batch_accum-001 | ckpt_batch_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINTED_BATCH_AND_ACCUM", "fix": {"grad_accum_steps": 2, "per_device_batch_size": 22}, "work": {"activation_per_sample_gb": 1.5, "max_per_device_batch": 22}} | {"archetype": "ckpt_batch_accum", "computations": ["R20", "R1", "R3"], "depth": 4, "interaction": "checkpointing x accum", "tier": "interacting"} |
train-checkpoint_retention-014 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 6}, "work": {"checkpoint_size_gb": 112.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-parallel_batch-007 | parallel_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PARALLEL_SHARDED_BATCH", "fix": {"per_device_batch_size": 64}, "work": {"fixed_memory_gb": 128.0, "params_per_rank_billions": 8.0}} | {"archetype": "parallel_batch", "computations": ["R22", "R13", "R1"], "depth": 4, "interaction": "parallelism x memory", "tier": "interacting"} |
train-shard_size-036 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 12500, "shard_size_gb": 40.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-accum_and_steps-036 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 1, "planned_total_steps": 2000}, "work": {"steps_per_epoch": 500}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_sampler_accum-030 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 8, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-shard_size-067 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 12500, "shard_size_gb": 25.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-parallel_batch-054 | parallel_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PARALLEL_SHARDED_BATCH", "fix": {"per_device_batch_size": 25}, "work": {"fixed_memory_gb": 64.0, "params_per_rank_billions": 4.0}} | {"archetype": "parallel_batch", "computations": ["R22", "R13", "R1"], "depth": 4, "interaction": "parallelism x memory", "tier": "interacting"} |
train-batch_from_components-005 | batch_from_components | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "BATCH_FROM_MEMORY_COMPONENTS", "fix": {"per_device_batch_size": 8}, "work": {"fixed_memory_gb": 64.0, "remaining_memory_gb": 16.0}} | {"archetype": "batch_from_components", "computations": ["R13", "R1"], "depth": 3, "interaction": null, "tier": "computed"} |
train-zero_batch-009 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 32}, "work": {"bytes_per_param_per_rank": 12.0, "fixed_memory_gb": 48.0, "remaining_memory_gb": 32.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-lr-003 | lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "LR_TOO_HIGH_NO_WARMUP", "fix": {"learning_rate": 2e-05, "warmup_ratio": 0.03}} | {"archetype": "lr", "computations": ["R4"], "depth": 1, "interaction": null, "tier": "diagnosis"} |
train-two_v100_dataloader-048 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-optimizer_batch-027 | optimizer_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "OPTIMIZER_MEMORY_BATCH", "fix": {"per_device_batch_size": 88}, "work": {"bytes_per_param": 10, "fixed_memory_gb": 40.0}} | {"archetype": "optimizer_batch", "computations": ["R21", "R1"], "depth": 3, "interaction": "optimizer x memory", "tier": "interacting"} |
train-accum_steps_lr-066 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 9.75e-05, "grad_accum_steps": 4, "planned_total_steps": 300}, "work": {"audit_step": 225, "steps_per_epoch": 100}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-two_sampler_accum-029 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 8, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-two_lora_vocab-045 | two_lora_vocab | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["ADAPTERS_FROZEN_ZERO_TRAINABLE", "EMBEDDING_VOCAB_MISMATCH"], "fix": {"adapter_requires_grad": true, "embedding_vocab_size": 50304}} | {"archetype": "two_lora_vocab", "computations": ["R8", "R11"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-shard_size-047 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 5000, "shard_size_gb": 25.0}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-zero_batch-048 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 20}, "work": {"bytes_per_param_per_rank": 11.0, "fixed_memory_gb": 88.0, "remaining_memory_gb": 40.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-two_oom_clipping-000 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 7}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-ckpt_batch_accum-041 | ckpt_batch_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINTED_BATCH_AND_ACCUM", "fix": {"grad_accum_steps": 8, "per_device_batch_size": 17}, "work": {"activation_per_sample_gb": 1.5, "max_per_device_batch": 17}} | {"archetype": "ckpt_batch_accum", "computations": ["R20", "R1", "R3"], "depth": 4, "interaction": "checkpointing x accum", "tier": "interacting"} |
train-optimizer_batch-015 | optimizer_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "OPTIMIZER_MEMORY_BATCH", "fix": {"per_device_batch_size": 44}, "work": {"bytes_per_param": 12, "fixed_memory_gb": 84.0}} | {"archetype": "optimizer_batch", "computations": ["R21", "R1"], "depth": 3, "interaction": "optimizer x memory", "tier": "interacting"} |
train-parallel_batch-001 | parallel_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PARALLEL_SHARDED_BATCH", "fix": {"per_device_batch_size": 85}, "work": {"fixed_memory_gb": 64.0, "params_per_rank_billions": 4.0}} | {"archetype": "parallel_batch", "computations": ["R22", "R13", "R1"], "depth": 4, "interaction": "parallelism x memory", "tier": "interacting"} |
train-trap_oom_checkpointing-008 | trap_oom_checkpointing | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACTIVATION_MEMORY_NO_CHECKPOINTING", "fix": {"gradient_checkpointing": true}} | {"archetype": "trap_oom_checkpointing", "computations": ["R1", "R2"], "depth": 2, "interaction": "oom_batch", "tier": "trap"} |
train-zero_batch-049 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 80}, "work": {"bytes_per_param_per_rank": 9.0, "fixed_memory_gb": 72.0, "remaining_memory_gb": 120.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-packing_time-026 | packing_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PACKED_TRAINING_TIME", "fix": {"planned_time_hours": 0.5}, "work": {"total_padded_tokens": 81920000, "total_real_tokens": 72000000}} | {"archetype": "packing_time", "computations": ["R23", "R16"], "depth": 4, "interaction": "packing x time", "tier": "interacting"} |
train-accum_steps_lr-015 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 9.75e-05, "grad_accum_steps": 8, "planned_total_steps": 500}, "work": {"audit_step": 375, "steps_per_epoch": 250}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-checkpoint_retention-006 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 4}, "work": {"checkpoint_size_gb": 112.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-checkpoint_retention-005 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 20}, "work": {"checkpoint_size_gb": 24.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-checkpoint_retention-020 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 4}, "work": {"checkpoint_size_gb": 48.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-bf16_v100-000 | bf16_v100 | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PRECISION_UNSUPPORTED_ON_GPU", "fix": {"dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "bf16_v100", "computations": ["R5", "R6"], "depth": 1, "interaction": null, "tier": "diagnosis"} |
train-shard_size-045 | shard_size | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "SHARD_SIZE_PER_RANK", "fix": {"examples_per_rank": 25000, "shard_size_gb": 12.5}} | {"archetype": "shard_size", "computations": ["R18"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_v100_dataloader-035 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-batch_from_components-050 | batch_from_components | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "BATCH_FROM_MEMORY_COMPONENTS", "fix": {"per_device_batch_size": 5}, "work": {"fixed_memory_gb": 32.0, "remaining_memory_gb": 8.0}} | {"archetype": "batch_from_components", "computations": ["R13", "R1"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_v100_dataloader-052 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-packing_time-030 | packing_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PACKED_TRAINING_TIME", "fix": {"planned_time_hours": 1.0}, "work": {"total_padded_tokens": 245760000, "total_real_tokens": 216000000}} | {"archetype": "packing_time", "computations": ["R23", "R16"], "depth": 4, "interaction": "packing x time", "tier": "interacting"} |
train-two_sampler_accum-024 | two_sampler_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["DUPLICATE_DATA_NO_DISTRIBUTED_SAMPLER", "EFFECTIVE_BATCH_MISMATCH"], "fix": {"grad_accum_steps": 2, "sampler": "DistributedSampler"}} | {"archetype": "two_sampler_accum", "computations": ["R7", "R3"], "depth": 3, "interaction": "sampler+accum", "tier": "two_fault"} |
train-training_time-068 | training_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "TRAINING_TIME_ESTIMATE", "fix": {"planned_time_hours": 0.17777777777777778}, "work": {"total_tokens": 102400000}} | {"archetype": "training_time", "computations": ["R16"], "depth": 3, "interaction": null, "tier": "computed"} |
train-accum_steps_lr-056 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 0.00011, "grad_accum_steps": 16, "planned_total_steps": 300}, "work": {"audit_step": 150, "steps_per_epoch": 100}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-packing_time-036 | packing_time | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "PACKED_TRAINING_TIME", "fix": {"planned_time_hours": 0.5}, "work": {"total_padded_tokens": 327680000, "total_real_tokens": 144000000}} | {"archetype": "packing_time", "computations": ["R23", "R16"], "depth": 4, "interaction": "packing x time", "tier": "interacting"} |
train-checkpoint_retention-021 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 10}, "work": {"checkpoint_size_gb": 48.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_lora_vocab-029 | two_lora_vocab | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["ADAPTERS_FROZEN_ZERO_TRAINABLE", "EMBEDDING_VOCAB_MISMATCH"], "fix": {"adapter_requires_grad": true, "embedding_vocab_size": 32064}} | {"archetype": "two_lora_vocab", "computations": ["R8", "R11"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-accum_and_steps-041 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 1, "planned_total_steps": 1000}, "work": {"steps_per_epoch": 250}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-dataloader-002 | dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "DATALOADER_FORK_WITH_CUDA", "fix": {"dataloader_multiprocessing_context": "spawn"}} | {"archetype": "dataloader", "computations": ["R9"], "depth": 1, "interaction": null, "tier": "diagnosis"} |
train-accum_steps_lr-038 | accum_steps_lr | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_STEPS_AND_LR", "fix": {"current_learning_rate": 0.000165, "grad_accum_steps": 2, "planned_total_steps": 600}, "work": {"audit_step": 300, "steps_per_epoch": 200}} | {"archetype": "accum_steps_lr", "computations": ["R3", "R14", "R15"], "depth": 5, "interaction": "accum x schedule", "tier": "interacting"} |
train-batch_from_components-068 | batch_from_components | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "BATCH_FROM_MEMORY_COMPONENTS", "fix": {"per_device_batch_size": 32}, "work": {"fixed_memory_gb": 48.0, "remaining_memory_gb": 32.0}} | {"archetype": "batch_from_components", "computations": ["R13", "R1"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_oom_clipping-024 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 12}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-checkpoint_retention-018 | checkpoint_retention | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINT_RETENTION_LIMIT", "fix": {"max_checkpoints_to_keep": 1}, "work": {"checkpoint_size_gb": 112.0}} | {"archetype": "checkpoint_retention", "computations": ["R17"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_oom_clipping-026 | two_oom_clipping | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["OOM_BATCH_TOO_LARGE", "NO_GRADIENT_CLIPPING"], "fix": {"max_grad_norm": 1.0, "per_device_batch_size": 18}} | {"archetype": "two_oom_clipping", "computations": ["R1", "R12"], "depth": 3, "interaction": "oom+clipping", "tier": "two_fault"} |
train-accum_and_steps-019 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 4, "planned_total_steps": 750}, "work": {"steps_per_epoch": 250}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-two_v100_dataloader-000 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-accum_and_steps-017 | accum_and_steps | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ACCUM_AND_TOTAL_STEPS", "fix": {"grad_accum_steps": 2, "planned_total_steps": 200}, "work": {"steps_per_epoch": 100}} | {"archetype": "accum_and_steps", "computations": ["R3", "R14"], "depth": 3, "interaction": null, "tier": "computed"} |
train-zero_batch-036 | zero_batch | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "ZERO_SHARDED_BATCH_SIZE", "fix": {"per_device_batch_size": 66}, "work": {"bytes_per_param_per_rank": 7.25, "fixed_memory_gb": 29.0, "remaining_memory_gb": 99.0}} | {"archetype": "zero_batch", "computations": ["R19", "R13", "R1"], "depth": 4, "interaction": "zero x memory", "tier": "interacting"} |
train-lr_at_step-016 | lr_at_step | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "LR_VALUE_AT_STEP", "fix": {"current_learning_rate": 0.000165}, "work": {"step_fraction": 0.5}} | {"archetype": "lr_at_step", "computations": ["R15"], "depth": 2, "interaction": null, "tier": "computed"} |
train-two_lora_vocab-022 | two_lora_vocab | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["ADAPTERS_FROZEN_ZERO_TRAINABLE", "EMBEDDING_VOCAB_MISMATCH"], "fix": {"adapter_requires_grad": true, "embedding_vocab_size": 32064}} | {"archetype": "two_lora_vocab", "computations": ["R8", "R11"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
train-ckpt_batch_accum-065 | ckpt_batch_accum | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": "CHECKPOINTED_BATCH_AND_ACCUM", "fix": {"grad_accum_steps": 4, "per_device_batch_size": 74}, "work": {"activation_per_sample_gb": 1.0, "max_per_device_batch": 74}} | {"archetype": "ckpt_batch_accum", "computations": ["R20", "R1", "R3"], "depth": 4, "interaction": "checkpointing x accum", "tier": "interacting"} |
train-two_v100_dataloader-007 | two_v100_dataloader | train | You are debugging a failed or misbehaving LLM training run.
Pick exactly ONE diagnosis code and give the minimal fix: only the config fields that must change, with corrected values.
Diagnosis codes:
- OOM_BATCH_TOO_LARGE
- ACTIVATION_MEMORY_NO_CHECKPOINTING
- EFFECTIVE_BATCH_MISMATCH
- LR_TOO_HIGH_NO_WARMUP
- PRECISIO... | {"diagnosis": ["PRECISION_UNSUPPORTED_ON_GPU", "DATALOADER_FORK_WITH_CUDA"], "fix": {"dataloader_multiprocessing_context": "spawn", "dtype": "fp16", "use_grad_scaler": true}} | {"archetype": "two_v100_dataloader", "computations": ["R5", "R9"], "depth": 2, "interaction": "mixed", "tier": "two_fault"} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.