MCPcopy Create free account

hub / github.com/AtomaAI/atoma-infer / functions

Functions613 in github.com/AtomaAI/atoma-infer

Methodload
( vb: VarBuilder, cfg: &Config, comm: Rc<Comm>, dtype: DType, device:
models/src/llama_nccl.rs:144
Methodload
( vb: VarBuilder, cfg: &MistralConfig, _dtype: DType, _device: &Device, )
models/src/mistral.rs:376
Methodload_multi
(vb: VarBuilder, prefixes: &[&str], comm: Rc<Comm>)
models/src/multi_gpu.rs:26
Methodlogit_bias
(&self)
server/src/api/chat_completions.rs:765
Methodlogprob
Getter for `logprob`
backends/vllm/src/sequence.rs:42
Functionmain
()
server/src/main.rs:30
Functionmain
()
csrc/build.rs:76
Functionmax_scale_exp2_sum
csrc/kernels/softmax.h:95
Methodmaybe_set_first_token_time
(&mut self, time: Instant)
backends/vllm/src/sequence.rs:1273
Methodn
(&self)
server/src/api/chat_completions.rs:781
Methodname
(&self)
models/src/multi_gpu.rs:67
Methodname
(&self)
csrc/src/ops.rs:26
Methodname
(&self)
csrc/src/lib.rs:346
Methodnew
( api_key: String, cache_dir: String, flush_storage: bool, model_name: String,
backends/vllm/src/config.rs:48
Methodnew
Constructor
backends/vllm/src/validation.rs:35
Methodnew
( n: usize, best_of: Option<usize>, presence_penalty: f32, frequency_penalty:
backends/vllm/src/sampling_params.rs:172
Methodnew
Constructor
backends/vllm/src/block_allocator.rs:29
Methodnew
Constructor
backends/vllm/src/evictor.rs:70
Methodnew
Creates a new `SchedulingBudget` with the specified token budget and maximum number of sequences.
backends/vllm/src/scheduler.rs:58
Methodnew
( cache_config: CacheConfig, device: Device, dtype: DType, model: M, e
backends/vllm/src/worker.rs:62
Methodnew
Constructor
backends/vllm/src/llm_engine.rs:61
Methodnew
Constructor
backends/vllm/src/sequence.rs:33
Methodnew
( block_size: usize, num_cpu_blocks: usize, num_gpu_blocks: usize, sliding_win
backends/vllm/src/block_manager.rs:55
Methodnew
Constructor
backends/vllm/src/block.rs:50
Methodnew
Creates a new `Streamer` with the specified receiver and model.
server/src/stream.rs:25
Methodnew
(dtype: DType, config: &Config, device: &Device)
models/src/llama.rs:155
Methodnew
( context_lengths: Tensor, slot_mapping: Tensor, query_start_locations: Tensor,
models/src/flash_attention.rs:89
Methodnew
(dtype: DType, cfg: &Phi3Config, dev: &Device)
models/src/phi3.rs:63
Methodnew
(linear: Linear)
models/src/multi_gpu.rs:13
Methodnew
(dtype: DType, cfg: &MistralConfig, dev: &Device)
models/src/mistral.rs:63
Methodnew_from_blocks
( block_size: usize, gpu_memory_utilization: f32, swap_space_fraction: f32, nu
backends/vllm/src/config.rs:229
Methodnormalize_softmax_lse
csrc/kernels/softmax.h:169
Methodnum_attention_heads
(&self)
backends/vllm/src/models/llama.rs:138
Methodnum_attention_heads
(&self)
backends/vllm/src/models/llama_nccl.rs:135
Methodnum_available_gpu_blocks
Determines the number of available GPU blocks
backends/vllm/src/worker.rs:105
Methodnum_blocks
(&self)
backends/vllm/src/evictor.rs:133
Methodnum_finished_sequences
Returns the number of finished sequences in the group. A finished sequence is one that has reached a terminal state (e.g., stopped, length capped, or
backends/vllm/src/sequence.rs:1830
Methodnum_gpu_block_override
Getter for `num_gpu_block_override`
backends/vllm/src/config.rs:310
Methodnum_hidden_layers
(&self)
backends/vllm/src/models/llama.rs:141
Methodnum_hidden_layers
(&self)
backends/vllm/src/models/llama_nccl.rs:138
Methodnum_key_value_heads
(&self)
models/src/phi3.rs:33
Methodnum_key_value_heads
(&self)
models/src/mistral.rs:33
Methodnum_kv_heads
(&self)
backends/vllm/src/models/llama.rs:144
Methodnum_kv_heads
(&self)
backends/vllm/src/models/llama_nccl.rs:141
Methodnum_unfinished_sequeces
Returns the total number of unfinished sequences across all queues. This method counts the sequences in the waiting, running, and swapped queues. #
backends/vllm/src/scheduler.rs:559
Methodoperator()
csrc/kernels/utils.h:91
Methodoperator()
csrc/kernels/utils.h:97
Methodoperator()
csrc/kernels/utils.h:104
Methodoutput_token_ids
Returns a clone of the output token IDs. # Returns A `Vec<u32>` containing the token IDs of the generated output.
backends/vllm/src/sequence.rs:451
Methodpoll_next
(mut self: Pin<&mut Self>, _cx: &mut Context<'_>)
server/src/stream.rs:71
Functionproper_tensor_print
(tensor: &Tensor, nested: usize)
help/src/lib.rs:5
Functionreduce_
csrc/kernels/softmax.h:47
Functionreduce_max
csrc/kernels/softmax.h:53
Functionreduce_sum
csrc/kernels/softmax.h:59
Methodremove
(&mut self, block_number: u32)
backends/vllm/src/evictor.rs:129
Methodrequest_id
Getter for `request_id`
backends/vllm/src/sequence.rs:2027
Functionrequest_schema_control
()
server/src/api/chat_completions.rs:1177
Functionreshape_and_cache_flash_t
Launches the `reshape_and_cache_kernel_flash` on the given `key_caches` and `value_caches`, respecting a slot mapping.
csrc/src/cache_manager.rs:349
Functionresolve_thread_kv_page_slice_offset
csrc/kernels/utils.h:297
Methodrun
csrc/kernels/utils.h:125
Functionrun_flash_fwd
csrc/kernels/flash_fwd_launch_template.h:61
Functionrun_flash_splitkv_fwd
csrc/kernels/flash_fwd_launch_template.h:108
Functionrun_mha_fwd_hdim128
csrc/kernels/flash_fwd_launch_template.h:235
Functionrun_mha_fwd_hdim160
csrc/kernels/flash_fwd_launch_template.h:273
Functionrun_mha_fwd_hdim192
csrc/kernels/flash_fwd_launch_template.h:304
Functionrun_mha_fwd_hdim224
csrc/kernels/flash_fwd_launch_template.h:321
Functionrun_mha_fwd_hdim256
csrc/kernels/flash_fwd_launch_template.h:348
Functionrun_mha_fwd_hdim32
csrc/kernels/flash_fwd_launch_template.h:180
Functionrun_mha_fwd_hdim64
csrc/kernels/flash_fwd_launch_template.h:188
Functionrun_mha_fwd_hdim96
csrc/kernels/flash_fwd_launch_template.h:208
Functionrun_mha_fwd_splitkv_dispatch
csrc/kernels/flash_fwd_launch_template.h:170
Methodsampling_strategy
Sampling strategy
backends/vllm/src/sampling_params.rs:410
Methodsequence_data
Returns a clone of the internal `SequenceData`. # Returns A cloned `SequenceData` instance. # Note This method performs a deep copy of the `Sequen
backends/vllm/src/sequence.rs:990
Methodsliding_window
(&self)
backends/vllm/src/tests/mod.rs:62
Methodsliding_window
(&self)
backends/vllm/src/models/llama.rs:147
Methodsliding_window
(&self)
backends/vllm/src/models/llama_nccl.rs:144
Methodsoftmax_rescale_o
csrc/kernels/softmax.h:136
Methodsoftmax_scale
(&self)
backends/vllm/src/models/llama.rs:150
Methodsoftmax_scale
(&self)
backends/vllm/src/models/llama_nccl.rs:147
Methodsort_by_priority
Sorts a collection of sequence groups by their priority. # Arguments `now` - The current time. `sequence_groups` - A queue of sequence groups to sor
backends/vllm/src/policy.rs:36
Methodstage
Returns the current processing stage of the sequence. # Returns The `SequenceStage` enum value representing the current stage (Prefill or Decode).
backends/vllm/src/sequence.rs:459
Methodstart
( service_request_receiver: UnboundedReceiver<ServiceRequest>, config_path: P, shutdow
backends/vllm/src/llm_service.rs:102
Methodstart
( config: M::C, devices_ids: Vec<usize>, dtype: DType, file_paths: ModelFilePa
backends/vllm/src/model_executor.rs:394
Methodstart
Starts the tokenizer workers
backends/vllm/src/tokenizer.rs:43
Methodsupported_head_sizes
Available supported head sizes
models/src/flash_attention.rs:233
Methodswap_space_bytes
Getter for `swap_space_bytes`
backends/vllm/src/config.rs:300
Functiontest_allocate
()
backends/vllm/src/block_manager.rs:1163
Functiontest_append_slot_single_seq
()
backends/vllm/src/block_manager.rs:1190
Functiontest_append_slot_with_cow
()
backends/vllm/src/block_manager.rs:1247
Functiontest_assistant_message
()
server/src/api/chat_completions.rs:1636
Functiontest_assistant_message_only
()
server/src/api/chat_completions.rs:1404
Functiontest_block_allocator_allocate
()
backends/vllm/src/block_allocator.rs:147
Functiontest_block_allocator_free
()
backends/vllm/src/block_allocator.rs:179
Functiontest_copy_blocks_bf16
()
csrc/tests/cache_manager_tests.rs:350
Functiontest_copy_blocks_different_dtypes
()
csrc/tests/cache_manager_tests.rs:493
Functiontest_copy_blocks_f16
()
csrc/tests/cache_manager_tests.rs:242
Functiontest_copy_blocks_invalid_block_mapping_shape
()
csrc/tests/cache_manager_tests.rs:525
Functiontest_copy_blocks_invalid_dtype
()
csrc/tests/cache_manager_tests.rs:509
Functiontest_copy_blocks_non_cuda_device
()
csrc/tests/cache_manager_tests.rs:477
Functiontest_copy_blocks_unequal_lengths
()
csrc/tests/cache_manager_tests.rs:458
← previousnext →401–500 of 613, ranked by callers