diff --git a/common/common.h b/common/common.h index 61b43cda42c7b..ee5826c75eeb6 100644 --- a/common/common.h +++ b/common/common.h @@ -174,7 +174,7 @@ struct common_params { float yarn_beta_fast = 32.0f; // YaRN low correction dim float yarn_beta_slow = 1.0f; // YaRN high correction dim int32_t yarn_orig_ctx = 0; // YaRN original context length - float defrag_thold = -1.0f; // KV cache defragmentation threshold + float defrag_thold = 0.1f; // KV cache defragmentation threshold struct cpu_params cpuparams; struct cpu_params cpuparams_batch; diff --git a/examples/chat-persistent.sh b/examples/chat-persistent.sh index d9cab9836482e..9d761ebb843af 100755 --- a/examples/chat-persistent.sh +++ b/examples/chat-persistent.sh @@ -23,8 +23,9 @@ CUR_PROMPT_CACHE="${CHAT_SAVE_DIR}/current-cache.bin" NEXT_PROMPT_FILE="${CHAT_SAVE_DIR}/next-prompt.txt" NEXT_PROMPT_CACHE="${CHAT_SAVE_DIR}/next-cache.bin" -SESSION_SIZE_MSG_PATTERN='main: session file matches [[:digit:]]+ / [[:digit:]]+' -SAMPLE_TIME_MSG_PATTERN='sample time =[[:space:]]+[[:digit:]]+.[[:digit:]]+ ms /[[:space:]]+[[:digit:]]+' +SESSION_AND_SAMPLE_PATTERN='main: session file matches [[:digit:]]+ / [[:digit:]]+'\ +'|'\ +'sampling time =[[:space:]]+[[:digit:]]+.[[:digit:]]+ ms /[[:space:]]+[[:digit:]]+' SED_DELETE_MESSAGES="/^(${USER_NAME}:|${AI_NAME}:|\\.\\.\\.)/,\$d" CTX_SIZE=2048 @@ -129,15 +130,12 @@ while read -e line; do printf ' ' - # HACK get num tokens from debug message - # TODO get both messages in one go - if ! session_size_msg="$(tail -n30 "$LOG" | grep -oE "$SESSION_SIZE_MSG_PATTERN")" || - ! sample_time_msg="$(tail -n10 "$LOG" | grep -oE "$SAMPLE_TIME_MSG_PATTERN")"; then + if ! session_and_sample_msg=$(tail -n30 "$LOG" | grep -oE "$SESSION_AND_SAMPLE_PATTERN"); then echo >&2 "Couldn't get number of tokens from ./llama-cli output!" exit 1 fi - n_tokens=$(($(cut -d/ -f2 <<<"$session_size_msg") + $(cut -d/ -f2 <<<"$sample_time_msg"))) + n_tokens=$(awk '{sum+=$1} END {print sum}' <<< "$(cut -d/ -f2 <<< "$session_and_sample_msg")") if ((n_tokens > CTX_ROTATE_POINT)); then tail -c+$((n_prompt_len_pre + 1)) "$CUR_PROMPT_FILE" >>"$NEXT_PROMPT_FILE" diff --git a/examples/convert_legacy_llama.py b/examples/convert_legacy_llama.py index 9ab9ab06edf8f..c4ec5c524e9b1 100755 --- a/examples/convert_legacy_llama.py +++ b/examples/convert_legacy_llama.py @@ -840,6 +840,8 @@ def add_meta_model(self, params: Params, metadata: gguf.Metadata | None) -> None self.gguf.add_base_model_version(key, base_model_entry["version"]) if "organization" in base_model_entry: self.gguf.add_base_model_organization(key, base_model_entry["organization"]) + if "description" in base_model_entry: + self.gguf.add_base_model_description(key, base_model_entry["description"]) if "url" in base_model_entry: self.gguf.add_base_model_url(key, base_model_entry["url"]) if "doi" in base_model_entry: @@ -849,12 +851,32 @@ def add_meta_model(self, params: Params, metadata: gguf.Metadata | None) -> None if "repo_url" in base_model_entry: self.gguf.add_base_model_repo_url(key, base_model_entry["repo_url"]) + if metadata.datasets is not None: + self.gguf.add_dataset_count(len(metadata.datasets)) + for key, dataset_entry in enumerate(metadata.datasets): + if "name" in dataset_entry: + self.gguf.add_dataset_name(key, dataset_entry["name"]) + if "author" in dataset_entry: + self.gguf.add_dataset_author(key, dataset_entry["author"]) + if "version" in dataset_entry: + self.gguf.add_dataset_version(key, dataset_entry["version"]) + if "organization" in dataset_entry: + self.gguf.add_dataset_organization(key, dataset_entry["organization"]) + if "description" in dataset_entry: + self.gguf.add_dataset_description(key, dataset_entry["description"]) + if "url" in dataset_entry: + self.gguf.add_dataset_url(key, dataset_entry["url"]) + if "doi" in dataset_entry: + self.gguf.add_dataset_doi(key, dataset_entry["doi"]) + if "uuid" in dataset_entry: + self.gguf.add_dataset_uuid(key, dataset_entry["uuid"]) + if "repo_url" in dataset_entry: + self.gguf.add_dataset_repo_url(key, dataset_entry["repo_url"]) + if metadata.tags is not None: self.gguf.add_tags(metadata.tags) if metadata.languages is not None: self.gguf.add_languages(metadata.languages) - if metadata.datasets is not None: - self.gguf.add_datasets(metadata.datasets) def add_meta_arch(self, params: Params) -> None: # Metadata About The Neural Architecture Itself diff --git a/examples/server/public/index.html b/examples/server/public/index.html index bf1d1b7940e8a..55639a9448e71 100644 --- a/examples/server/public/index.html +++ b/examples/server/public/index.html @@ -200,23 +200,38 @@

Settings

System Message
-