Не попадитесь на накрученные каналы! Узнайте, не накручивает ли канал просмотры или
подписчиков
Проверить канал на накрутку
Телеграм канал «1000iq»
Подписчики
Всего
72
Сегодня
0
Просмотров на пост
Всего
44
ER
Общий
60.8%
Динамика публикаций
Telemetr - сервис глубокой аналитики
телеграм-каналов
телеграм-каналов
Получите подробную информацию о каждом канале
Отберите самые эффективные каналы для
рекламных размещений, по приросту подписчиков,
ER, количеству просмотров на пост и другим метрикам
рекламных размещений, по приросту подписчиков,
ER, количеству просмотров на пост и другим метрикам
Анализируйте рекламные посты
и креативы
и креативы
Узнайте какие посты лучше сработали,
а какие хуже, даже если их давно удалили
а какие хуже, даже если их давно удалили
Оценивайте эффективность тематики и контента
Узнайте, какую тематику лучше не рекламировать
на канале, а какая зайдет на ура
на канале, а какая зайдет на ура
Показано 1 из 25 постов
Смотреть все посты
Пост от 04.10.2026 00:23
3
0
0
======================================================================
DATASET
======================================================================
Contracts found: 2
======================================================================
[1/2] Reentrance
======================================================================
[transformers] causal_conv1d_fn is falling back to its reference PyTorch implementation because causal_conv1d is not installed. This is correct but much slower; install causal_conv1d for the optimized kernel.
[transformers] chunk_gated_delta_rule is falling back to its reference PyTorch implementation because flash-linear-attention is not installed. This is correct but much slower; install flash-linear-attention for the optimized kernel.
[transformers] causal_conv1d_update is falling back to its reference PyTorch implementation because causal_conv1d is not installed. This is correct but much slower; install causal_conv1d for the optimized kernel.
[transformers] fused_recurrent_gated_delta_rule is falling back to its reference PyTorch implementation because flash-linear-attention is not installed. This is correct but much slower; install flash-linear-attention for the optimized kernel.
ROUGE-1: 0.385
ROUGE-2: 0.108
ROUGE-L: 0.144
Reference severity: critical
Generated severity: critical
Severity match: True
Repetition loop: False
Generated words: 375
======================================================================
[2/2] Switch
======================================================================
ROUGE-1: 0.446
ROUGE-2: 0.112
ROUGE-L: 0.164
Reference severity: medium
Generated severity: critical
Severity match: False
Repetition loop: False
Generated words: 629
======================================================================
МОДЕЛЬ: Qwen/Qwen3.5-9b
thinking=on
4bit=True
LoRA=on
LoRA adapter=./adapter
======================================================================
--- Reentrance (n=1) ---
ROUGE-1 F: 0.385
ROUGE-2 F: 0.108
ROUGE-L F: 0.144
Severity accuracy (top): 100.0% (1 примеров с severity)
Зацикливаний: 0/1 (0.0%)
Средняя длина ответа: 375 слов
--- Switch (n=1) ---
ROUGE-1 F: 0.446
ROUGE-2 F: 0.112
ROUGE-L F: 0.164
Severity accuracy (top): 0.0% (1 примеров с severity)
Зацикливаний: 0/1 (0.0%)
Средняя длина ответа: 629 слов
--- ОБЩЕЕ (n=2) ---
ROUGE-1 F: 0.415
ROUGE-2 F: 0.110
ROUGE-L F: 0.154
Severity accuracy (top): 50.0% (2 примеров с severity)
Зацикливаний: 0/2 (0.0%)
Средняя длина ответа: 502 слов
======================================================================
Evaluation finished.
======================================================================
DATASET
======================================================================
Contracts found: 2
======================================================================
[1/2] Reentrance
======================================================================
[transformers] causal_conv1d_fn is falling back to its reference PyTorch implementation because causal_conv1d is not installed. This is correct but much slower; install causal_conv1d for the optimized kernel.
[transformers] chunk_gated_delta_rule is falling back to its reference PyTorch implementation because flash-linear-attention is not installed. This is correct but much slower; install flash-linear-attention for the optimized kernel.
[transformers] causal_conv1d_update is falling back to its reference PyTorch implementation because causal_conv1d is not installed. This is correct but much slower; install causal_conv1d for the optimized kernel.
[transformers] fused_recurrent_gated_delta_rule is falling back to its reference PyTorch implementation because flash-linear-attention is not installed. This is correct but much slower; install flash-linear-attention for the optimized kernel.
ROUGE-1: 0.385
ROUGE-2: 0.108
ROUGE-L: 0.144
Reference severity: critical
Generated severity: critical
Severity match: True
Repetition loop: False
Generated words: 375
======================================================================
[2/2] Switch
======================================================================
ROUGE-1: 0.446
ROUGE-2: 0.112
ROUGE-L: 0.164
Reference severity: medium
Generated severity: critical
Severity match: False
Repetition loop: False
Generated words: 629
======================================================================
МОДЕЛЬ: Qwen/Qwen3.5-9b
thinking=on
4bit=True
LoRA=on
LoRA adapter=./adapter
======================================================================
--- Reentrance (n=1) ---
ROUGE-1 F: 0.385
ROUGE-2 F: 0.108
ROUGE-L F: 0.144
Severity accuracy (top): 100.0% (1 примеров с severity)
Зацикливаний: 0/1 (0.0%)
Средняя длина ответа: 375 слов
--- Switch (n=1) ---
ROUGE-1 F: 0.446
ROUGE-2 F: 0.112
ROUGE-L F: 0.164
Severity accuracy (top): 0.0% (1 примеров с severity)
Зацикливаний: 0/1 (0.0%)
Средняя длина ответа: 629 слов
--- ОБЩЕЕ (n=2) ---
ROUGE-1 F: 0.415
ROUGE-2 F: 0.110
ROUGE-L F: 0.154
Severity accuracy (top): 50.0% (2 примеров с severity)
Зацикливаний: 0/2 (0.0%)
Средняя длина ответа: 502 слов
======================================================================
Evaluation finished.
======================================================================