Datasets:

LunaNguyenX
/

Pretrain_language_model-1BL3-competesmoe4

Modalities:

Formats:

Size:

Libraries:

Dataset card Data Studio Files Files and versions

Dataset Viewer

Auto-converted to Parquet Duplicate

Split (1)

train · 1 rows

profile null	name string	save_dir string	reset int64	log string	save_interval string	wandb_save_interval string	seed string	gpu string	keep_alive int64	sweep_id_for_grid_search int64	restore string	wandb_bug_workaround int64	wandb_sync_checkpoints int64	batch_size int64	lr float64	min_lr_multiplier float64	wd float64	lr_warmup int64	test_interval int64	n_microbatch string	per_device_batch_size string	lr_sched.steps string	lr_sched.gamma float64	lr_sched.type string	length_bucketed_sampling int64	grad_clip string	test_batch_size string	val_log_details int64	reg_scales string	reg_lin_decay string	reg float64	optimizer string	adam.betas string	adam.eps float64	stop_after string	amp int64	bfloat16 int64	nan_detect int64	max_length_per_batch string	log_grad_norms int64	speedtest string	dump_logs int64	debug_plot_interval string	lm.trafo.context_blocks int64	lm.trafo.test_context_blocks string	lm.trafo.same_length_eval int64	lm.trafo.same_length int64	lm.trafo.last_layer_context int64	lm.trafo.xl_init int64	lm.trafo.norm_input int64	rope.rotate_fraction float64	rope.base float64	pkm.n_heads int64	moe.n_experts int64	moe.expert_size int64	moe_name string	moe.selection_mode string	moe.perplexity_reg float64	moe.perplexity_reg_mode string	moe.att.perplexity_reg_mode string	moe.activation_after_topk int64	moe.att.expert_size int64	moe.topk int64	moe.bias int64	moe.sel_bias int64	moe.dropout_factor float64	moe.drop_expert float64	moe.sync_distributed int64	moe.init_scale float64	moe.att.n_experts int64	moe.att.enable int64	moe.att.q_expert int64	moe.att.k_expert int64	moe.att.v_expert int64	moe.att.o_expert int64	moe.att.k int64	moe.att.v_size string	moe.att.same_sel int64	moe.att.expert_dropout string	moe.att.selection_mode string	moe.att.perplexity_reg string	moe.att.drop_expert string	moe.att.separate_kq_sel int64	moe.att.norm_init int64	moe.att.dropout float64	moe.att.selection_dropout float64	moe.nonorm int64	in_topk int64	balance_affinity int64	is_cosine int64	is_norm_weight int64	norm_softmax int64	norm_sigmoid int64	moa.cvloss float64	moa.switchloss float64	moa.zloss float64	balance_loss_coef float64	balance_loss_coef_comp float64	router_z_loss_coef float64	router_loss_coef float64	max_compete_in_iter int64	warm_up float64	rate_flip float64	router_theta float64	scale_weight float64	hybrid int64	tribrid int64	moa.miloss float64	sut.sample_topk int64	sut.max_relative_positions int64	sut.drop_gate float64	moe.selection_dropout float64	moe.layer_std_constant float64	transformer.universal.group_size int64	transformer.universal.group_type string	transformer.embedding_scale string	transformer.topk_value int64	transformer.activation string	transformer.p_drop_layer float64	transformer.head_projection_size string	transformer.act_loss float64	transformer.plot_head_details int64	lm.trafo.force_out_norm int64	plot.n_steps int64	dump_validation_plots string	details_log_interval string	lm.state_drop_probability float64	lm.unroll int64	lm.unroll_eval string	lm.example_context int64	lm.example_window int64	lm.eval.blimp.batch_mul int64	lm.eval.enabled int64	lm.eval.lambada.enabled int64	lm.eval.cbt.batch_mul int64	lm.eval.cbt.length_limit string	lm.eval.cbt.enabled int64	lm.eval.cbt.end_only int64	lm.eval.blimp.enabled int64	lm.eval.hellaswag.enabled int64	lm.eval.hellaswag.batch_mul int64	lm.eval.piqa.enabled int64	lm.eval.piqa.batch_mul int64	lm.eval.ai2arc.enabled int64	lm.eval.ai2arc.batch_mul int64	lm.eval.mmlu.enabled int64	lm.eval.openbookqa.enabled int64	lm.eval.race.enabled int64	lm.eval.siqa.enabled int64	lm.eval.winogrande.enabled int64	lm.eval.commonsenseqa.enabled int64	sentencepiece.n_pieces int64	lmds.valid_ratio float64	thestack.languages string	state_size int64	task string	dropout float64	embedding_size string	transformer.n_heads int64	transformer.variant string	transformer.ff_multiplier float64	transformer.encoder_n_layers int64	transformer.attention_dropout float64	load_pretrained_model null	test_pretrained int64	train_baseline int64	test_only int64	save_name_logs string	fs_cache_pattern string
null	slimpajama_moe_no_attmoe_1B_competesmoe	/cm/archive/namnv78_new/revise_checkpoints/pretrain_final/1BL3/competesmoe	0	tb	10000	None	none	auto	0	0	/cm/archive/namnv78_new/revise_checkpoints/pretrain_final/1BL3/competesmoe/slimpajama_moe_no_attmoe_1B_competesmoe/checkpoint/model-10000.pth	0	0	64	0.00025	0.1	0.01	4,000	1,000	None	16		0.1	cos	0	0.25	None	0			1	adamw	0.9,0.999	0	200000	1	1	0	none	0	none	0	none	0	none	0	0	0	0	0	0.5	10,000	8	24	512	competesmoe	gate	0.01	step	none	0	256	2	0	0	1	0	1	1	4	0	1	1	1	1	2	none	0	none	sigmoid	none	none	0	0	0	0	0	0	1	0	0	0	0	0	0	0	0.01	0.01	0.001	0.001	3	0.05	0.07	0.2	1	1	0	0	0	64	0	0	2	24	abab	none	32	relu	0	128	0	0	0	-128		500	0	1,024	none	100	40	16	0	0	1	none	0	0	0	0	16	0	16	0	4	0	0	0	0	0	0	8,000	0.005	python,html,c++,rust,javascript,haskell,scala,assembly	1,024	slimpajama_transformer	0	none	32	preln_moe	2	24	0	null	1	0	0	results	*

No dataset card yet

Downloads last month: 3

Number of rows:

Total file size: