Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/AtomaAI/atoma-infer
/ functions
Functions
613 in github.com/AtomaAI/atoma-infer
⨍
Functions
613
◇
Types & classes
173
Method
load
( vb: VarBuilder, cfg: &Config, comm: Rc<Comm>, dtype: DType, device:
models/src/llama_nccl.rs:144
Method
load
( vb: VarBuilder, cfg: &MistralConfig, _dtype: DType, _device: &Device, )
models/src/mistral.rs:376
Method
load_multi
(vb: VarBuilder, prefixes: &[&str], comm: Rc<Comm>)
models/src/multi_gpu.rs:26
Method
logit_bias
(&self)
server/src/api/chat_completions.rs:765
Method
logprob
Getter for `logprob`
backends/vllm/src/sequence.rs:42
Function
main
()
server/src/main.rs:30
Function
main
()
csrc/build.rs:76
Function
max_scale_exp2_sum
csrc/kernels/softmax.h:95
Method
maybe_set_first_token_time
(&mut self, time: Instant)
backends/vllm/src/sequence.rs:1273
Method
n
(&self)
server/src/api/chat_completions.rs:781
Method
name
(&self)
models/src/multi_gpu.rs:67
Method
name
(&self)
csrc/src/ops.rs:26
Method
name
(&self)
csrc/src/lib.rs:346
Method
new
( api_key: String, cache_dir: String, flush_storage: bool, model_name: String,
backends/vllm/src/config.rs:48
Method
new
Constructor
backends/vllm/src/validation.rs:35
Method
new
( n: usize, best_of: Option<usize>, presence_penalty: f32, frequency_penalty:
backends/vllm/src/sampling_params.rs:172
Method
new
Constructor
backends/vllm/src/block_allocator.rs:29
Method
new
Constructor
backends/vllm/src/evictor.rs:70
Method
new
Creates a new `SchedulingBudget` with the specified token budget and maximum number of sequences.
backends/vllm/src/scheduler.rs:58
Method
new
( cache_config: CacheConfig, device: Device, dtype: DType, model: M, e
backends/vllm/src/worker.rs:62
Method
new
Constructor
backends/vllm/src/llm_engine.rs:61
Method
new
Constructor
backends/vllm/src/sequence.rs:33
Method
new
( block_size: usize, num_cpu_blocks: usize, num_gpu_blocks: usize, sliding_win
backends/vllm/src/block_manager.rs:55
Method
new
Constructor
backends/vllm/src/block.rs:50
Method
new
Creates a new `Streamer` with the specified receiver and model.
server/src/stream.rs:25
Method
new
(dtype: DType, config: &Config, device: &Device)
models/src/llama.rs:155
Method
new
( context_lengths: Tensor, slot_mapping: Tensor, query_start_locations: Tensor,
models/src/flash_attention.rs:89
Method
new
(dtype: DType, cfg: &Phi3Config, dev: &Device)
models/src/phi3.rs:63
Method
new
(linear: Linear)
models/src/multi_gpu.rs:13
Method
new
(dtype: DType, cfg: &MistralConfig, dev: &Device)
models/src/mistral.rs:63
Method
new_from_blocks
( block_size: usize, gpu_memory_utilization: f32, swap_space_fraction: f32, nu
backends/vllm/src/config.rs:229
Method
normalize_softmax_lse
csrc/kernels/softmax.h:169
Method
num_attention_heads
(&self)
backends/vllm/src/models/llama.rs:138
Method
num_attention_heads
(&self)
backends/vllm/src/models/llama_nccl.rs:135
Method
num_available_gpu_blocks
Determines the number of available GPU blocks
backends/vllm/src/worker.rs:105
Method
num_blocks
(&self)
backends/vllm/src/evictor.rs:133
Method
num_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
Method
num_gpu_block_override
Getter for `num_gpu_block_override`
backends/vllm/src/config.rs:310
Method
num_hidden_layers
(&self)
backends/vllm/src/models/llama.rs:141
Method
num_hidden_layers
(&self)
backends/vllm/src/models/llama_nccl.rs:138
Method
num_key_value_heads
(&self)
models/src/phi3.rs:33
Method
num_key_value_heads
(&self)
models/src/mistral.rs:33
Method
num_kv_heads
(&self)
backends/vllm/src/models/llama.rs:144
Method
num_kv_heads
(&self)
backends/vllm/src/models/llama_nccl.rs:141
Method
num_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
Method
operator()
csrc/kernels/utils.h:91
Method
operator()
csrc/kernels/utils.h:97
Method
operator()
csrc/kernels/utils.h:104
Method
output_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
Method
poll_next
(mut self: Pin<&mut Self>, _cx: &mut Context<'_>)
server/src/stream.rs:71
Function
proper_tensor_print
(tensor: &Tensor, nested: usize)
help/src/lib.rs:5
Function
reduce_
csrc/kernels/softmax.h:47
Function
reduce_max
csrc/kernels/softmax.h:53
Function
reduce_sum
csrc/kernels/softmax.h:59
Method
remove
(&mut self, block_number: u32)
backends/vllm/src/evictor.rs:129
Method
request_id
Getter for `request_id`
backends/vllm/src/sequence.rs:2027
Function
request_schema_control
()
server/src/api/chat_completions.rs:1177
Function
reshape_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
Function
resolve_thread_kv_page_slice_offset
csrc/kernels/utils.h:297
Method
run
csrc/kernels/utils.h:125
Function
run_flash_fwd
csrc/kernels/flash_fwd_launch_template.h:61
Function
run_flash_splitkv_fwd
csrc/kernels/flash_fwd_launch_template.h:108
Function
run_mha_fwd_hdim128
csrc/kernels/flash_fwd_launch_template.h:235
Function
run_mha_fwd_hdim160
csrc/kernels/flash_fwd_launch_template.h:273
Function
run_mha_fwd_hdim192
csrc/kernels/flash_fwd_launch_template.h:304
Function
run_mha_fwd_hdim224
csrc/kernels/flash_fwd_launch_template.h:321
Function
run_mha_fwd_hdim256
csrc/kernels/flash_fwd_launch_template.h:348
Function
run_mha_fwd_hdim32
csrc/kernels/flash_fwd_launch_template.h:180
Function
run_mha_fwd_hdim64
csrc/kernels/flash_fwd_launch_template.h:188
Function
run_mha_fwd_hdim96
csrc/kernels/flash_fwd_launch_template.h:208
Function
run_mha_fwd_splitkv_dispatch
csrc/kernels/flash_fwd_launch_template.h:170
Method
sampling_strategy
Sampling strategy
backends/vllm/src/sampling_params.rs:410
Method
sequence_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
Method
sliding_window
(&self)
backends/vllm/src/tests/mod.rs:62
Method
sliding_window
(&self)
backends/vllm/src/models/llama.rs:147
Method
sliding_window
(&self)
backends/vllm/src/models/llama_nccl.rs:144
Method
softmax_rescale_o
csrc/kernels/softmax.h:136
Method
softmax_scale
(&self)
backends/vllm/src/models/llama.rs:150
Method
softmax_scale
(&self)
backends/vllm/src/models/llama_nccl.rs:147
Method
sort_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
Method
stage
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
Method
start
( service_request_receiver: UnboundedReceiver<ServiceRequest>, config_path: P, shutdow
backends/vllm/src/llm_service.rs:102
Method
start
( config: M::C, devices_ids: Vec<usize>, dtype: DType, file_paths: ModelFilePa
backends/vllm/src/model_executor.rs:394
Method
start
Starts the tokenizer workers
backends/vllm/src/tokenizer.rs:43
Method
supported_head_sizes
Available supported head sizes
models/src/flash_attention.rs:233
Method
swap_space_bytes
Getter for `swap_space_bytes`
backends/vllm/src/config.rs:300
Function
test_allocate
()
backends/vllm/src/block_manager.rs:1163
Function
test_append_slot_single_seq
()
backends/vllm/src/block_manager.rs:1190
Function
test_append_slot_with_cow
()
backends/vllm/src/block_manager.rs:1247
Function
test_assistant_message
()
server/src/api/chat_completions.rs:1636
Function
test_assistant_message_only
()
server/src/api/chat_completions.rs:1404
Function
test_block_allocator_allocate
()
backends/vllm/src/block_allocator.rs:147
Function
test_block_allocator_free
()
backends/vllm/src/block_allocator.rs:179
Function
test_copy_blocks_bf16
()
csrc/tests/cache_manager_tests.rs:350
Function
test_copy_blocks_different_dtypes
()
csrc/tests/cache_manager_tests.rs:493
Function
test_copy_blocks_f16
()
csrc/tests/cache_manager_tests.rs:242
Function
test_copy_blocks_invalid_block_mapping_shape
()
csrc/tests/cache_manager_tests.rs:525
Function
test_copy_blocks_invalid_dtype
()
csrc/tests/cache_manager_tests.rs:509
Function
test_copy_blocks_non_cuda_device
()
csrc/tests/cache_manager_tests.rs:477
Function
test_copy_blocks_unequal_lengths
()
csrc/tests/cache_manager_tests.rs:458
← previous
next →
401–500 of 613, ranked by callers