Rdoublea's workspace
Runs
1
Name
1 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
batch_size
checkpointer._component_
checkpointer.checkpoint_dir
checkpointer.checkpoint_files
checkpointer.model_type
checkpointer.output_dir
compile
dataset
dataset._component_
device
dtype
enable_activation_checkpointing
epochs
gradient_accumulation_steps
log_every_n_steps
log_peak_memory_stats
loss._component_
metric_logger._component_
metric_logger.name
metric_logger.project
model._component_
optimizer._component_
optimizer.fused
optimizer.lr
output_dir
resume_from_checkpoint
shuffle
tokenizer._component_
tokenizer.max_seq_len
tokenizer.path
global_step
loss
lr
peak_memory_active
peak_memory_alloc
peak_memory_reserved
tokens_per_second_per_gpu
Finished
Add notes...
rdoublea
1h 5m 22s
-
4
torchtune.training.FullModelHFCheckpointer
/tmp/Llama-3.2-1B-Instruct/
["model.safetensors"]
LLAMA3_2
/tmp/Llama-3.2-1B-Instruct/
true
[{"_component_":"dataset.evol_code_alpaca"},{"_component_":"dataset.flytech_python_codes"},{"_component_":"dataset.python_code_instructions_alpaca"},{"_component_":"dataset.code_feedback_filtered_instruction"}]
-
cuda
bf16
false
5
4
1
true
torchtune.modules.loss.CEWithChunkedOutputLoss
torchtune.training.metric_logging.WandBLogger
llama3.2-1B
coding-agent
torchtune.models.llama3_2.llama3_2_1b
torch.optim.AdamW
true
0.0001
/tmp/full-llama3.2-finetune
false
true
torchtune.models.llama3.llama3_tokenizer
8192
/tmp/Llama-3.2-1B-Instruct/original/tokenizer.model
620
0.59666
0.0001
51.18728
51.18728
53.76563
23914.82511
1-1
of 1