Skip to content

Remove dead ASR task and fix object-detection UI conditionals #174

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Alvin-Nahabwe wants to merge 1 commit into
base: fix/dl-client-contract
Choose a base branch
from
Open

Remove dead ASR task and fix object-detection UI conditionals #174

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
152 changes: 6 additions & 146 deletions server/deep_learning.R
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,6 @@ deep_learning = function() {

# --- For Inference Tab ---
obj_inference_result <- reactiveVal(list(status = "Ready", image_url = NULL, error = NULL))
asr_inference_result <- reactiveVal(list(status = "Ready", transcription = NULL, error = NULL))
img_class_inference_result <- reactiveVal(list(status = "Ready", prediction = NULL, error = NULL))
seg_inference_result <- reactiveVal(list(status = "Ready", image_url = NULL, error = NULL))

Expand Down Expand Up @@ -76,13 +75,11 @@ deep_learning = function() {
observe({
task <- input$task_selector
if (task == "object_detection") {
shinyjs::show("obj_panel"); shinyjs::hide("asr_panel"); shinyjs::hide("img_class_panel"); shinyjs::hide("seg_panel")
} else if (task == "asr") {
shinyjs::hide("obj_panel"); shinyjs::show("asr_panel"); shinyjs::hide("img_class_panel"); shinyjs::hide("seg_panel")
shinyjs::show("obj_panel"); shinyjs::hide("img_class_panel"); shinyjs::hide("seg_panel")
} else if (task == "image_classification") {
shinyjs::hide("obj_panel"); shinyjs::hide("asr_panel"); shinyjs::show("img_class_panel"); shinyjs::hide("seg_panel")
shinyjs::hide("obj_panel"); shinyjs::show("img_class_panel"); shinyjs::hide("seg_panel")
} else if (task == "image_segmentation") {
shinyjs::hide("obj_panel"); shinyjs::hide("asr_panel"); shinyjs::hide("img_class_panel"); shinyjs::show("seg_panel")
shinyjs::hide("obj_panel"); shinyjs::hide("img_class_panel"); shinyjs::show("seg_panel")
}
})

Expand All @@ -96,9 +93,6 @@ deep_learning = function() {
if (task_slug == "object_detection") {
arch_choices <- names(model_registry()$object_detection)
updateSelectInput(session, "obj_model_arch", choices = arch_choices)
} else if (task_slug == "asr") {
arch_choices <- names(model_registry()$asr)
updateSelectInput(session, "asr_model_arch", choices = arch_choices)
} else if (task_slug == "image_classification") {
arch_choices <- names(model_registry()$image_classification)
updateSelectInput(session, "img_class_model_arch", choices = arch_choices)
Expand All @@ -114,11 +108,6 @@ deep_learning = function() {
checkpoints <- model_registry()$object_detection[[input$obj_model_arch]]
updateSelectInput(session, "obj_model_checkpoint", choices = checkpoints)
})
observeEvent(input$asr_model_arch, {
req(model_registry(), input$asr_model_arch, input$asr_model_arch != "Loading...")
checkpoints <- model_registry()$asr[[input$asr_model_arch]]
updateSelectInput(session, "asr_model_checkpoint", choices = checkpoints)
})
observeEvent(input$img_class_model_arch, {
req(model_registry(), input$img_class_model_arch, input$img_class_model_arch != "Loading...")
checkpoints <- model_registry()$image_classification[[input$img_class_model_arch]]
Expand Down Expand Up @@ -156,8 +145,6 @@ deep_learning = function() {
task_slug <- input$task_selector
if (task_slug == "object_detection") {
updateSelectInput(session, "obj_dataset_id", choices = load_datasets_for_task("object_detection"))
} else if (task_slug == "asr") {
updateSelectInput(session, "asr_dataset_id", choices = load_datasets_for_task("asr"))
} else if (task_slug == "image_classification") {
updateSelectInput(session, "img_class_dataset_id", choices = load_datasets_for_task("image_classification"))
} else if (task_slug == "image_segmentation") {
Expand Down Expand Up @@ -235,7 +222,7 @@ deep_learning = function() {
refresh_data_trigger() # React to the trigger

tryCatch({
tasks <- c("object_detection", "asr", "image_classification", "image_segmentation")
tasks <- c("object_detection", "image_classification", "image_segmentation")
all_datasets <- lapply(tasks, function(task) {
req <- request(paste0(api_url, "/data/list/", task))
resp_data <- resp_body_json(req_perform(req), simplifyVector = TRUE)
Expand Down Expand Up @@ -321,80 +308,7 @@ deep_learning = function() {
})
})

# --- 4.2: ASR Job ---
observeEvent(input$start_asr_job, {
req(input$asr_dataset_id, input$asr_model_checkpoint)
reset_live_training_ui("ASR")

outlier_val <- input$outlier_std_devs
if (is.null(outlier_val) || is.na(outlier_val) || !is.numeric(outlier_val) || !input$asr_apply_outlier_filtering) {
outlier_val <- 2.0
}

max_hours <- if (is.na(input$asr_max_train_hours) || is.null(input$asr_max_train_hours)) NULL else as.character(input$asr_max_train_hours)

tryCatch({
req_list <- list(
dataset_id = as.character(input$asr_dataset_id),
model_checkpoint = as.character(input$asr_model_checkpoint),
run_name = as.character(input$asr_run_name),
version = as.character(input$asr_version),
language = as.character(input$asr_language),
language_code = as.character(input$asr_language_code),
speaker_id_column = as.character(input$asr_speaker_id_column),
text_column = as.character(input$asr_text_column),
target_sampling_rate = as.character(input$asr_target_sampling_rate),
min_duration_s = as.character(input$asr_min_duration_s),
max_duration_s = as.character(input$asr_max_duration_s),
min_transcript_len = as.character(input$asr_min_transcript_len),
max_transcript_len = as.character(input$asr_max_transcript_len),
apply_outlier_filtering = as.character(input$asr_apply_outlier_filtering),
outlier_std_devs = as.character(outlier_val),
is_presplit = as.character(input$asr_is_presplit),
speaker_disjointness = as.character(input$asr_speaker_disjointness),
train_ratio = as.character(input$asr_train_ratio),
dev_ratio = as.character(input$asr_dev_ratio),
test_ratio = as.character(input$asr_test_ratio),
epochs = as.character(input$asr_epochs),
learning_rate = as.character(input$asr_learning_rate),
lr_scheduler_type = as.character(input$asr_lr_scheduler_type),
warmup_ratio = as.character(input$asr_warmup_ratio),
train_batch_size = as.character(input$asr_train_batch_size),
eval_batch_size = as.character(input$asr_eval_batch_size),
gradient_accumulation_steps = as.character(input$asr_gradient_accumulation_steps),
gradient_checkpointing = as.character(input$asr_gradient_checkpointing),
optimizer = as.character(input$asr_optimizer),
early_stopping_patience = as.character(input$asr_early_stopping_patience),
early_stopping_threshold = as.character(input$asr_early_stopping_threshold),
push_to_hub = as.character(input$asr_push_to_hub),
hub_user_id = as.character(input$asr_hub_user_id),
hub_private_repo = as.character(input$asr_hub_private_repo),
log_to_wandb = as.character(input$asr_log_to_wandb),
wandb_project = as.character(input$asr_wandb_project),
wandb_entity = as.character(input$asr_wandb_entity),
seed = as.character(input$asr_seed),
num_proc = as.character(input$asr_num_proc),
max_train_hours = max_hours
)

req_list <- req_list[!sapply(req_list, is.null)]

req <- request(paste0(api_url, "/train/asr")) %>%
req_body_multipart(!!!req_list)

resp <- req_perform(req)
resp_data <- resp_body_json(resp)
active_job_id(resp_data$job_id)
polled_data(list(status = "Queued", task = "ASR", log = "Job is queued."))

}, error = function(e) {
error_message <- as.character(e$message)
if(!is.null(e$body)) { error_message <- paste("API Error:", e$body) }
polled_data(list(status = "Error", task = "ASR", log = error_message))
})
})

# --- 4.3: Image Classification Job ---
# --- 4.2: Image Classification Job ---
observeEvent(input$start_img_class_job, {
req(input$img_class_dataset_id, input$img_class_model_checkpoint)
reset_live_training_ui("Image Classification")
Expand Down Expand Up @@ -435,7 +349,7 @@ deep_learning = function() {
})
})

# --- 4.4: Image Segmentation Job ---
# --- 4.3: Image Segmentation Job ---
observeEvent(input$start_seg_job, {
req(input$seg_dataset_id, input$seg_model_checkpoint)
reset_live_training_ui("Image Segmentation")
Expand Down Expand Up @@ -906,23 +820,6 @@ deep_learning = function() {
}
})

observeEvent(input$infer_asr_run_name, {
run_name <- input$infer_asr_run_name
if (nchar(run_name) > 2) {
tryCatch({
req <- request(paste0(api_url, "/checkpoints")) %>%
req_url_query(run_name = run_name, task_type = "asr")
resp <- req_perform(req)
if (resp_status(resp) == 200) {
checkpoints <- resp_body_json(resp, simplifyVector = TRUE)
updateSelectInput(session, "infer_asr_checkpoint_dropdown", choices = checkpoints)
}
}, error = function(e) {
updateSelectInput(session, "infer_asr_checkpoint_dropdown", choices = c("Error finding checkpoints"))
})
}
})

observeEvent(input$infer_img_class_run_name, {
run_name <- input$infer_img_class_run_name
if (nchar(run_name) > 2) {
Expand Down Expand Up @@ -971,26 +868,6 @@ deep_learning = function() {
})
})

observeEvent(input$start_asr_inference, {
req(input$infer_asr_audio_upload)
req(input$infer_asr_checkpoint_dropdown)
asr_inference_result(list(status = "Running...", transcription = "Processing...", error = NULL))
tryCatch({
req <- request(paste0(api_url, "/inference/asr")) %>%
req_body_multipart(
audio = curl::form_file(input$infer_asr_audio_upload$datapath),
model_checkpoint = input$infer_asr_checkpoint_dropdown
)
resp <- req_perform(req)
resp_data <- resp_body_json(resp)
asr_inference_result(list(status = "Success", transcription = resp_data$transcription, error = NULL))
}, error = function(e) {
error_message <- as.character(e$message)
if(!is.null(e$body)) { error_message <- paste("API Error:", e$body) }
asr_inference_result(list(status = "Error", transcription = NULL, error = error_message))
})
})

observeEvent(input$start_img_class_inference, {
req(input$infer_img_class_upload, input$infer_img_class_checkpoint_dropdown)
img_class_inference_result(list(status = "Running...", prediction = "Processing...", error = NULL))
Expand Down Expand Up @@ -1045,23 +922,6 @@ deep_learning = function() {
list(src = temp_file, contentType = 'image/jpeg', alt = "Inference Result")
}, deleteFile = TRUE)

output$asr_inference_status_ui <- renderUI({
res <- asr_inference_result()
if (res$status == "Running...") {
tags$div(class = "alert alert-info", "Running inference...")
} else if (res$status == "Error") {
tags$div(class = "alert alert-danger", HTML(paste("<strong>Error:</strong>", res$error)))
}
})
output$asr_transcription_output <- renderText({
res <- asr_inference_result()
if (is.null(res$transcription)) {
"Upload an audio file and click 'Run Inference' to see the transcription here."
} else {
res$transcription
}
})

output$img_class_inference_status_ui <- renderUI({
res <- img_class_inference_result()
if (res$status == "Running...") tags$div(class = "alert alert-info", "Running inference...")
Expand Down
Loading