Functions — github.com/ESHyperscale/HyperscaleES

hub / github.com/ESHyperscale/HyperscaleES / functions

Functions274 in github.com/ESHyperscale/HyperscaleES

⨍ Functions 274◇ Types & classes 60

↓ 83 callersFunctioncall_submodule

(cls, name, common_params, *args, **kwargs)

src/hyperscalees/models/common.py:58

↓ 16 callersMethod__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps)

src/hyperscalees/environments/llm_bandits.py:141

↓ 11 callersMethodencode

(self, src)

src/hyperscalees/models/llm/tokenizer.py:106

↓ 10 callersMethodrand_init

(cls, key, n_embd, space_gymnax, dtype, *args, **kwargs)

src/hyperscalees/models/rl.py:36

↓ 9 callersMethodget_input

(self, indices)

src/hyperscalees/environments/llm_bandits.py:156

↓ 9 callersFunctionmerge_inits

(**kwargs)

src/hyperscalees/models/common.py:37

↓ 8 callersFunctionbuild_generate_thread

(MODEL, NOISER, frozen_noiser_params, config, base_evo_keys, master_gen_key, temperature=1.0)

llm_experiments/utils.py:31

↓ 8 callersMethoddecode

(self, tokens)

src/hyperscalees/models/llm/tokenizer.py:109

↓ 8 callersMethodforward

Forward pass of model returns just the output

src/hyperscalees/models/base_model.py:29

↓ 8 callersMethodget_batch_fitness

(self, indices, full_generations)

src/hyperscalees/environments/llm_bandits.py:159

↓ 8 callersMethodinit_noiser

Return frozen_noiser_params and noiser_params

src/hyperscalees/noiser/sparse.py:117

↓ 6 callersMethoddefault_state

(cls, params, config)

src/hyperscalees/models/llm/llm.py:75

↓ 6 callersFunctionget_model

(model_name, dtype=None, rwkv_type="BaseRWKV", verbose=False, reload_cache=False)

src/hyperscalees/models/llm/auto.py:95

↓ 6 callersFunctionsafe_decode

(tokens, tokenizer)

llm_experiments/utils.py:22

↓ 6 callersFunctionsimple_es_tree_key

(params, base_key, scan_map)

src/hyperscalees/models/common.py:31

↓ 5 callersFunctionas_named

(x, mesh, spec)

llm_experiments/utils.py:16

↓ 5 callersFunctionget_padded_prompt

(single_prompt, generation_length)

src/hyperscalees/environments/llm_bandits.py:55

↓ 5 callersMethodrand_init

(cls, key, in_dim, out_dim, dtype, *args, **kwargs)

src/hyperscalees/models/common.py:85

↓ 5 callersFunctionreplicate_matrix

(x)

llm_experiments/general_do_evolution_multi_gpu.py:194

↓ 5 callersFunctionsafe_decode

Decode up to the first 0 token, else full sequence.

src/hyperscalees/environments/llm_bandits.py:44

↓ 4 callersFunction_to_cpu_tree

Bring a (possibly sharded/replicated) pytree to host numpy on proc 0.

llm_experiments/sft_evolution.py:302

↓ 4 callersFunction_to_cpu_tree

Bring a (possibly sharded/replicated) pytree to host numpy on proc 0.

llm_experiments/general_do_evolution_multi_gpu.py:306

↓ 4 callersMethodconvert_fitnesses

(cls, frozen_noiser_params, noiser_params, raw_scores, num_episodes_list=None)

src/hyperscalees/noiser/sparse.py:164

↓ 4 callersMethoddo_updates

(cls, frozen_noiser_params, noiser_params, params, base_keys, fitnesses, iterinfos, es_map)

src/hyperscalees/noiser/sparse.py:189

↓ 4 callersFunctionreplicate_matrix

(x)

llm_experiments/do_grpo_multi_gpu.py:144

↓ 4 callersFunctionreplicate_matrix

(x)

llm_experiments/sft_evolution.py:178

↓ 3 callersFunctionget_nonlora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/open_es.py:8

↓ 3 callersFunctionsave

Save the Any model as a file given a path. See https://github.com/google/jax/issues/2116#issuecomment-580322624 :param model: The Any m

src/hyperscalees/models/llm/auto.py:134

↓ 3 callersFunctionstrip_thoughts

(txt: str)

src/hyperscalees/environments/llm_bandits.py:60

↓ 2 callersMethod_do_update

(cls, param, base_key, fitnesses, iterinfos, map_classification, sigma, frozen_noiser_params, **kwargs)

src/hyperscalees/noiser/eggroll.py:117

↓ 2 callersFunction_get_aime_answer

(example)

src/hyperscalees/environments/llm_bandits.py:325

↓ 2 callersFunction_score_with_math_verify

(pred_text, gold_text)

src/hyperscalees/environments/llm_bandits.py:329

↓ 2 callersFunction_to_camel

(s: str)

src/hyperscalees/environments/llm_bandits.py:448

↓ 2 callersFunction_to_jnp_tree

(x, dtype_str: Optional[str])

llm_experiments/general_do_evolution_multi_gpu.py:140

↓ 2 callersFunctionbatch_calculate_fitnesses

(x)

tests/end_to_end_test.py:46

↓ 2 callersFunctionbuild_validate

(MODEL, config, params_example, base_evo_keys, master_gen_key, tokenizer, legacy_tokenizer, args, temperature=

llm_experiments/utils.py:101

↓ 2 callersFunctionfold_in_helper

(key, epoch, true_thread_idx)

llm_experiments/utils.py:19

↓ 2 callersFunctionforward_and_sample

(noiser_params, params, input_token, input_state, generation_key, iterinfo)

llm_experiments/utils.py:33

↓ 2 callersFunctionget_lora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/eggroll_baseline_subtraction.py:30

↓ 2 callersFunctionget_lora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/eggroll.py:11

↓ 2 callersFunctionget_lora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/alteggroll.py:10

↓ 2 callersFunctionget_space_defn

(space)

src/hyperscalees/models/rl.py:11

↓ 2 callersFunctionget_sparse_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/sparse.py:12

↓ 2 callersFunctionload

Read the Any model from a file See https://github.com/google/jax/issues/2116#issuecomment-580322624 :param path: The path to read the m

src/hyperscalees/models/llm/auto.py:156

↓ 2 callersFunctionreplicate_matrix

(x)

llm_experiments/do_grpo.py:141

↓ 2 callersFunctionreplicate_matrix

(x)

llm_experiments/general_do_evolution.py:154

↓ 2 callersFunctionvalidate

(params, epoch)

llm_experiments/utils.py:118

↓ 1 callersMethod_do_update

(cls, param, base_key, fitnesses, iterinfos, map_classification, sigma, frozen_noiser_params, **kwargs)

src/hyperscalees/noiser/sparse.py:177

↓ 1 callersMethod_do_update

(cls, param, base_key, fitnesses, iterinfos, map_classification, sigma, frozen_noiser_params, **kwargs)

src/hyperscalees/noiser/open_es.py:95

↓ 1 callersMethod_do_update

(cls, param, base_key, fitnesses, iterinfos, map_classification, sigma, frozen_noiser_params, **kwargs)

src/hyperscalees/noiser/eggroll_baseline_subtraction.py:138

↓ 1 callersMethod_do_update

(cls, param, base_key, fitnesses, iterinfos, map_classification, sigma, frozen_noiser_params, **kwargs)

src/hyperscalees/noiser/alteggroll.py:116

↓ 1 callersMethod_do_updates_batched

This performs the update, but does so in a clever batched way, so that we do not end up having a python for loop over over every single weight matrix.

src/hyperscalees/noiser/eggroll.py:142

↓ 1 callersMethod_do_updates_original

(cls, frozen_noiser_params, noiser_params, params, base_keys, fitnesses, iterinfos, es_map)

src/hyperscalees/noiser/eggroll.py:136

↓ 1 callersMethod_forward

(cls, common_params, *args, **kwargs)

src/hyperscalees/models/base_model.py:40

↓ 1 callersMethod_forward

(cls, common_params, x, *args, **kwargs)

src/hyperscalees/models/common.py:94

↓ 1 callersFunction_make_rg_train_class

(dataset_name: str)

src/hyperscalees/environments/llm_bandits.py:452

↓ 1 callersFunction_make_rg_validation_class

(dataset_name: str)

src/hyperscalees/environments/llm_bandits.py:473

↓ 1 callersFunction_parse_gold_cached

(gold_text: str)

src/hyperscalees/environments/llm_bandits.py:12

↓ 1 callersFunction_resolve_ckpt_path

(load_path: str | Path)

llm_experiments/general_do_evolution_multi_gpu.py:127

↓ 1 callersFunctionbuild_generate_sft_thread

(MODEL, NOISER, frozen_noiser_params, config, base_evo_keys, master_gen_key, temperature=1.0)

llm_experiments/utils.py:64

↓ 1 callersFunctioncheck_accuracy

(generated_ans, solution)

src/hyperscalees/environments/llm_bandits.py:196

↓ 1 callersMethoddecodeBytes

(self, tokens)

src/hyperscalees/models/llm/tokenizer.py:67

↓ 1 callersMethoddo_Tmm

(cls, frozen_noiser_params, noiser_params, param, base_key, iterinfo, x)

src/hyperscalees/noiser/sparse.py:142

↓ 1 callersMethoddo_emb

(cls, frozen_noiser_params, noiser_params, param, base_key, iterinfo, x)

src/hyperscalees/noiser/sparse.py:153

↓ 1 callersMethoddo_mm

(cls, frozen_noiser_params, noiser_params, param, base_key, iterinfo, x)

src/hyperscalees/noiser/sparse.py:131

↓ 1 callersMethodembed

(cls, common_params, tokens)

src/hyperscalees/models/llm/llm.py:79

↓ 1 callersMethodencode

(self, src)

src/hyperscalees/models/llm/tokenizer.py:87

↓ 1 callersMethodencodeBytes

(self, src: bytes)

src/hyperscalees/models/llm/tokenizer.py:46

↓ 1 callersFunctionextract_ground_truth

(text)

src/hyperscalees/environments/llm_bandits.py:192

↓ 1 callersFunctionextract_predicted_answer

(text)

src/hyperscalees/environments/llm_bandits.py:175

↓ 1 callersMethodforward_seq

(cls, common_params, x, state, length, new_starts)

src/hyperscalees/models/llm/llm.py:87

↓ 1 callersMethodget_es_map

(cls, config)

src/hyperscalees/models/llm/llm.py:23

↓ 1 callersMethodget_noisy_standard

(cls, frozen_noiser_params, noiser_params, param, base_key, iterinfo)

src/hyperscalees/noiser/sparse.py:158

↓ 1 callersFunctionget_nonlora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/sparse.py:57

↓ 1 callersFunctionget_nonlora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/eggroll_baseline_subtraction.py:14

↓ 1 callersFunctionget_nonlora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/eggroll.py:27

↓ 1 callersFunctionget_nonlora_update_params

(frozen_noiser_params, base_sigma, iterinfo, param, key)

src/hyperscalees/noiser/alteggroll.py:26

↓ 1 callersMethodget_scan_map

(cls, config)

src/hyperscalees/models/llm/llm.py:19

↓ 1 callersFunctionlayer_norm

(x, eps=1e-5)

src/hyperscalees/models/common.py:11

↓ 1 callersMethodload_from_torch

Loads from torch (cpu) model and outputs a jax (cpu) model

src/hyperscalees/models/llm/llm.py:27

↓ 1 callersFunctionmake_rg_prompt

(question: str)

src/hyperscalees/environments/llm_bandits.py:65

↓ 1 callersFunctionmerge_frozen

(common, **kwargs)

src/hyperscalees/models/common.py:53

↓ 1 callersMethodouthead

(cls, common_params, x)

src/hyperscalees/models/llm/llm.py:83

↓ 1 callersFunctionsingle_epoch

(optimizer, noiser_params, params, true_train_fitness_sum, epoch)

llm_experiments/do_grpo.py:278

↓ 1 callersFunctionsingle_epoch

(optimizer, noiser_params, params, true_train_fitness_sum, epoch)

llm_experiments/do_grpo_multi_gpu.py:368

↓ 1 callersFunctionsingle_epoch

(noiser_params, params, true_train_fitness_sum, epoch)

llm_experiments/general_do_evolution.py:230

↓ 1 callersFunctionsingle_epoch

(noiser_params, params, true_train_fitness_sum, epoch)

llm_experiments/sft_evolution.py:381

↓ 1 callersFunctionsingle_epoch

(noiser_params, params, true_train_fitness_sum, epoch)

llm_experiments/general_do_evolution_multi_gpu.py:367

↓ 1 callersFunctionsingle_fitness

(generated_answer, true_answer, i)

src/hyperscalees/environments/llm_bandits.py:202

↓ 1 callersFunctiontile_state_for_batch

(state_single, batch)

llm_experiments/sft_evolution.py:378

↓ 1 callersMethodtransform_config

(cls, config)

src/hyperscalees/models/llm/llm.py:15

↓ 1 callersMethodtransform_torch_model

(cls, torch_model, dtype=jnp.bfloat16)

src/hyperscalees/models/llm/llm.py:11

↓ 1 callersFunctionvalidate

Runs validation over args.validation_iterations * total_validation_generations samples.

llm_experiments/do_grpo_multi_gpu.py:323

↓ 1 callersFunctionvalidate

Runs validation over args.validation_iterations * total_validation_generations samples.

llm_experiments/sft_evolution.py:315

↓ 1 callersFunctionvalidate

Runs validation over args.validation_iterations * total_validation_generations samples.

llm_experiments/general_do_evolution_multi_gpu.py:319

Method__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps)

src/hyperscalees/environments/llm_bandits.py:77

Method__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps, single_fitness)

src/hyperscalees/environments/llm_bandits.py:96

Method__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps)

src/hyperscalees/environments/llm_bandits.py:112

Method__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps)

src/hyperscalees/environments/llm_bandits.py:126

Method__init__

(self, encoding_tokenizer, decoding_tokenizer, max_num_steps)

src/hyperscalees/environments/llm_bandits.py:211

next →1–100 of 274, ranked by callers