Upload dataset
Browse files
1xA100_EXPERTS_5_QUANT_SBMM_MULTILAUNCH/train-00000-of-00001.parquet
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8ac984ef3d819d68cea7cef0d7cb5cf2fb978bff03c56bf2b9760843d37f378b
|
3 |
+
size 56508
|
README.md
CHANGED
@@ -3144,6 +3144,268 @@ dataset_info:
|
|
3144 |
num_examples: 7
|
3145 |
download_size: 56283
|
3146 |
dataset_size: 9732
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3147 |
- config_name: 1xA100_EXPERTS_5_UNQUANT
|
3148 |
features:
|
3149 |
- name: server_args
|
@@ -4241,6 +4503,10 @@ configs:
|
|
4241 |
data_files:
|
4242 |
- split: train
|
4243 |
path: 1xA100_EXPERTS_5_QUANT_NAIVE/train-*
|
|
|
|
|
|
|
|
|
4244 |
- config_name: 1xA100_EXPERTS_5_UNQUANT
|
4245 |
data_files:
|
4246 |
- split: train
|
|
|
3144 |
num_examples: 7
|
3145 |
download_size: 56283
|
3146 |
dataset_size: 9732
|
3147 |
+
- config_name: 1xA100_EXPERTS_5_QUANT_SBMM_MULTILAUNCH
|
3148 |
+
features:
|
3149 |
+
- name: server_args
|
3150 |
+
struct:
|
3151 |
+
- name: allow_toppings_registration
|
3152 |
+
dtype: bool
|
3153 |
+
- name: api_key
|
3154 |
+
dtype: 'null'
|
3155 |
+
- name: attention_backend
|
3156 |
+
dtype: string
|
3157 |
+
- name: base_min_new_token_ratio
|
3158 |
+
dtype: float64
|
3159 |
+
- name: chat_template
|
3160 |
+
dtype: 'null'
|
3161 |
+
- name: chunked_prefill_size
|
3162 |
+
dtype: int64
|
3163 |
+
- name: constrained_json_whitespace_pattern
|
3164 |
+
dtype: 'null'
|
3165 |
+
- name: context_length
|
3166 |
+
dtype: int64
|
3167 |
+
- name: controller_port
|
3168 |
+
dtype: int64
|
3169 |
+
- name: debug
|
3170 |
+
dtype: bool
|
3171 |
+
- name: decode_log_interval
|
3172 |
+
dtype: int64
|
3173 |
+
- name: detokenizer_ipc_name
|
3174 |
+
dtype: string
|
3175 |
+
- name: detokenizer_port
|
3176 |
+
dtype: int64
|
3177 |
+
- name: device
|
3178 |
+
dtype: string
|
3179 |
+
- name: disable_cuda_graph
|
3180 |
+
dtype: bool
|
3181 |
+
- name: disable_cuda_graph_padding
|
3182 |
+
dtype: bool
|
3183 |
+
- name: disable_custom_all_reduce
|
3184 |
+
dtype: bool
|
3185 |
+
- name: disable_disk_cache
|
3186 |
+
dtype: bool
|
3187 |
+
- name: disable_flashinfer
|
3188 |
+
dtype: bool
|
3189 |
+
- name: disable_flashinfer_sampling
|
3190 |
+
dtype: bool
|
3191 |
+
- name: disable_mla
|
3192 |
+
dtype: bool
|
3193 |
+
- name: disable_nan_detection
|
3194 |
+
dtype: bool
|
3195 |
+
- name: disable_penalizer
|
3196 |
+
dtype: bool
|
3197 |
+
- name: disable_radix_cache
|
3198 |
+
dtype: bool
|
3199 |
+
- name: disable_regex_jump_forward
|
3200 |
+
dtype: bool
|
3201 |
+
- name: dist_init_addr
|
3202 |
+
dtype: 'null'
|
3203 |
+
- name: dp_size
|
3204 |
+
dtype: int64
|
3205 |
+
- name: dtype
|
3206 |
+
dtype: string
|
3207 |
+
- name: enable_double_sparsity
|
3208 |
+
dtype: bool
|
3209 |
+
- name: enable_dp_attention
|
3210 |
+
dtype: bool
|
3211 |
+
- name: enable_mixed_chunk
|
3212 |
+
dtype: bool
|
3213 |
+
- name: enable_overlap_schedule
|
3214 |
+
dtype: bool
|
3215 |
+
- name: enable_p2p_check
|
3216 |
+
dtype: bool
|
3217 |
+
- name: enable_stats_logging
|
3218 |
+
dtype: bool
|
3219 |
+
- name: enable_system_controller
|
3220 |
+
dtype: bool
|
3221 |
+
- name: enable_toppings
|
3222 |
+
dtype: bool
|
3223 |
+
- name: enable_torch_compile
|
3224 |
+
dtype: bool
|
3225 |
+
- name: flashinfer_workspace_size
|
3226 |
+
dtype: int64
|
3227 |
+
- name: host
|
3228 |
+
dtype: string
|
3229 |
+
- name: init_new_token_ratio
|
3230 |
+
dtype: float64
|
3231 |
+
- name: init_number_of_deltas
|
3232 |
+
dtype: int64
|
3233 |
+
- name: init_number_of_loras
|
3234 |
+
dtype: int64
|
3235 |
+
- name: init_toppings
|
3236 |
+
dtype: 'null'
|
3237 |
+
- name: is_embedding
|
3238 |
+
dtype: bool
|
3239 |
+
- name: kv_cache_dtype
|
3240 |
+
dtype: string
|
3241 |
+
- name: load_balance_method
|
3242 |
+
dtype: string
|
3243 |
+
- name: load_format
|
3244 |
+
dtype: string
|
3245 |
+
- name: log_requests
|
3246 |
+
dtype: bool
|
3247 |
+
- name: lora_paths
|
3248 |
+
dtype: 'null'
|
3249 |
+
- name: max_lora_ranks
|
3250 |
+
dtype: int64
|
3251 |
+
- name: max_prefill_tokens
|
3252 |
+
dtype: int64
|
3253 |
+
- name: max_running_requests
|
3254 |
+
dtype: 'null'
|
3255 |
+
- name: max_toppings_per_batch
|
3256 |
+
dtype: int64
|
3257 |
+
- name: max_torch_compile_bs
|
3258 |
+
dtype: int64
|
3259 |
+
- name: max_total_tokens
|
3260 |
+
dtype: 'null'
|
3261 |
+
- name: mem_fraction_static
|
3262 |
+
dtype: float64
|
3263 |
+
- name: model_path
|
3264 |
+
dtype: string
|
3265 |
+
- name: nccl_ports
|
3266 |
+
sequence: string
|
3267 |
+
- name: new_token_ratio_decay
|
3268 |
+
dtype: float64
|
3269 |
+
- name: nnodes
|
3270 |
+
dtype: int64
|
3271 |
+
- name: node_rank
|
3272 |
+
dtype: int64
|
3273 |
+
- name: num_continue_decode_steps
|
3274 |
+
dtype: int64
|
3275 |
+
- name: num_continuous_decode_steps
|
3276 |
+
dtype: int64
|
3277 |
+
- name: port
|
3278 |
+
dtype: int64
|
3279 |
+
- name: quantization
|
3280 |
+
dtype: 'null'
|
3281 |
+
- name: random_seed
|
3282 |
+
dtype: int64
|
3283 |
+
- name: retract_decode_steps
|
3284 |
+
dtype: int64
|
3285 |
+
- name: sampling_backend
|
3286 |
+
dtype: string
|
3287 |
+
- name: schedule_conservativeness
|
3288 |
+
dtype: float64
|
3289 |
+
- name: schedule_policy
|
3290 |
+
dtype: string
|
3291 |
+
- name: scheduler_input_ipc_name
|
3292 |
+
dtype: string
|
3293 |
+
- name: scheduler_port
|
3294 |
+
dtype: int64
|
3295 |
+
- name: served_model_name
|
3296 |
+
dtype: string
|
3297 |
+
- name: server_id
|
3298 |
+
dtype: string
|
3299 |
+
- name: show_time_cost
|
3300 |
+
dtype: bool
|
3301 |
+
- name: skip_tokenizer_init
|
3302 |
+
dtype: bool
|
3303 |
+
- name: stream_interval
|
3304 |
+
dtype: int64
|
3305 |
+
- name: tokenizer_ipc_name
|
3306 |
+
dtype: string
|
3307 |
+
- name: tokenizer_mode
|
3308 |
+
dtype: string
|
3309 |
+
- name: tokenizer_path
|
3310 |
+
dtype: string
|
3311 |
+
- name: tokenizer_port
|
3312 |
+
dtype: int64
|
3313 |
+
- name: torchao_config
|
3314 |
+
dtype: string
|
3315 |
+
- name: tp_size
|
3316 |
+
dtype: int64
|
3317 |
+
- name: triton_attention_reduce_in_fp32
|
3318 |
+
dtype: bool
|
3319 |
+
- name: trust_remote_code
|
3320 |
+
dtype: bool
|
3321 |
+
- name: use_heterogeneous_pool
|
3322 |
+
dtype: bool
|
3323 |
+
- name: watchdog_timeout
|
3324 |
+
dtype: float64
|
3325 |
+
- name: client_args
|
3326 |
+
struct:
|
3327 |
+
- name: dataset
|
3328 |
+
dtype: string
|
3329 |
+
- name: endpoint
|
3330 |
+
dtype: string
|
3331 |
+
- name: goodput
|
3332 |
+
dtype: 'null'
|
3333 |
+
- name: metric_percentiles
|
3334 |
+
dtype: string
|
3335 |
+
- name: model
|
3336 |
+
dtype: string
|
3337 |
+
- name: num_prompts
|
3338 |
+
dtype: int64
|
3339 |
+
- name: output
|
3340 |
+
dtype: string
|
3341 |
+
- name: percentile_metrics
|
3342 |
+
dtype: string
|
3343 |
+
- name: request_rate
|
3344 |
+
dtype: float64
|
3345 |
+
- name: tokenizer
|
3346 |
+
dtype: string
|
3347 |
+
- name: wait_until_ready
|
3348 |
+
dtype: bool
|
3349 |
+
- name: metrics
|
3350 |
+
struct:
|
3351 |
+
- name: completed
|
3352 |
+
dtype: int64
|
3353 |
+
- name: mean_e2el_ms
|
3354 |
+
dtype: float64
|
3355 |
+
- name: mean_itl_ms
|
3356 |
+
dtype: float64
|
3357 |
+
- name: mean_tpot_ms
|
3358 |
+
dtype: float64
|
3359 |
+
- name: mean_ttft_ms
|
3360 |
+
dtype: float64
|
3361 |
+
- name: median_e2el_ms
|
3362 |
+
dtype: float64
|
3363 |
+
- name: median_itl_ms
|
3364 |
+
dtype: float64
|
3365 |
+
- name: median_tpot_ms
|
3366 |
+
dtype: float64
|
3367 |
+
- name: median_ttft_ms
|
3368 |
+
dtype: float64
|
3369 |
+
- name: output_throughput
|
3370 |
+
dtype: float64
|
3371 |
+
- name: percentiles_e2el_ms
|
3372 |
+
sequence:
|
3373 |
+
sequence: float64
|
3374 |
+
- name: percentiles_itl_ms
|
3375 |
+
sequence:
|
3376 |
+
sequence: float64
|
3377 |
+
- name: percentiles_tpot_ms
|
3378 |
+
sequence:
|
3379 |
+
sequence: float64
|
3380 |
+
- name: percentiles_ttft_ms
|
3381 |
+
sequence:
|
3382 |
+
sequence: float64
|
3383 |
+
- name: request_goodput
|
3384 |
+
dtype: float64
|
3385 |
+
- name: request_throughput
|
3386 |
+
dtype: float64
|
3387 |
+
- name: std_e2el_ms
|
3388 |
+
dtype: float64
|
3389 |
+
- name: std_itl_ms
|
3390 |
+
dtype: float64
|
3391 |
+
- name: std_tpot_ms
|
3392 |
+
dtype: float64
|
3393 |
+
- name: std_ttft_ms
|
3394 |
+
dtype: float64
|
3395 |
+
- name: total_input
|
3396 |
+
dtype: int64
|
3397 |
+
- name: total_output
|
3398 |
+
dtype: int64
|
3399 |
+
- name: total_token_throughput
|
3400 |
+
dtype: float64
|
3401 |
+
- name: device
|
3402 |
+
dtype: string
|
3403 |
+
splits:
|
3404 |
+
- name: train
|
3405 |
+
num_bytes: 9998
|
3406 |
+
num_examples: 7
|
3407 |
+
download_size: 56508
|
3408 |
+
dataset_size: 9998
|
3409 |
- config_name: 1xA100_EXPERTS_5_UNQUANT
|
3410 |
features:
|
3411 |
- name: server_args
|
|
|
4503 |
data_files:
|
4504 |
- split: train
|
4505 |
path: 1xA100_EXPERTS_5_QUANT_NAIVE/train-*
|
4506 |
+
- config_name: 1xA100_EXPERTS_5_QUANT_SBMM_MULTILAUNCH
|
4507 |
+
data_files:
|
4508 |
+
- split: train
|
4509 |
+
path: 1xA100_EXPERTS_5_QUANT_SBMM_MULTILAUNCH/train-*
|
4510 |
- config_name: 1xA100_EXPERTS_5_UNQUANT
|
4511 |
data_files:
|
4512 |
- split: train
|