========== YENİ KOMUT | 16.09.2026 22:43:34 ==========
========== YENİ KOMUT | 16.09.2026 22:43:34 ==========
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/compute.alpha.json | hits=2834 =====
138: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
173: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
248: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
278: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
314: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
349: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
403: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
485: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
513: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
550: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
598: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
641: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
716: "description": "Advise how, where and when to create the requested amount of instances\nwith specified accelerators, within the specified time and location limits.\nThe method recommends creating future reservations for
752: "capacity": {
753: "description": "Advice on making real-time decisions (such as choosing zone or\nmachine types) during deployment to maximize your chances of obtaining\ncapacity.",
754: "flatPath": "projects/{project}/regions/{region}/advice/capacity",
756: "id": "compute.advice.capacity",
777: "path": "projects/{project}/regions/{region}/advice/capacity",
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/compute.beta.json | hits=2523 =====
138: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
173: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
248: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
278: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
314: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
349: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
403: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
485: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
513: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
550: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
598: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
641: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
716: "description": "Advise how, where and when to create the requested amount of instances\nwith specified accelerators, within the specified time and location limits.\nThe method recommends creating future reservations for
766: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
801: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
848: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
930: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
965: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/compute.v1.json | hits=2331 =====
138: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
173: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
248: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
278: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
314: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
349: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
403: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
485: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
513: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
550: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
598: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
641: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
716: "description": "Advise how, where and when to create the requested amount of instances\nwith specified accelerators, within the specified time and location limits.\nThe method recommends creating future reservations for
766: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
801: "description": "Opt-in for partial success behavior which provides partial results in case\nof failure. The default value is false.\n\nFor example, when partial success behavior is enabled, aggregatedList for a\nsingle z
848: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
930: "description": "An optional request ID to identify requests. Specify a unique request ID so\nthat if you must retry your request, the server will know to ignore the\nrequest if it has already been completed.\n\nFor examp
965: "description": "A filter expression that filters resources listed in the response. Most\nCompute resources support two types of filter expressions:\nexpressions that support regular expressions and expressions that follo
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/telegram/ext/_extbot.py | hits=1953 =====
140: from telegram.ext import BaseRateLimiter, Defaults
155: All API methods of this class have an additional keyword argument ``rate_limit_args``.
156: This can be used to pass additional information to the rate limiter, specifically to
157: :paramref:`telegram.ext.BaseRateLimiter.process_request.rate_limit_args`.
160: the generic type of the :attr:`rate_limiter` used by the bot. Use :obj:`None` if no rate
161: limiter is used.
164: * The keyword argument ``rate_limit_args`` can `not` be used, if :attr:`rate_limiter`
167: additional argument, as this method will never be rate limited.
193: rate_limiter (:class:`telegram.ext.BaseRateLimiter`, optional): A rate limiter to use for
194: limiting the number of requests made by the bot per time interval.
200: __slots__ = ("_callback_data_cache", "_defaults", "_rate_limiter")
235: rate_limiter: "BaseRateLimiter[RLARGS] | None" = None,
250: rate_limiter: "BaseRateLimiter[RLARGS] | None" = None,
264: self._rate_limiter: BaseRateLimiter | None = rate_limiter
318: :paramref:`ExtBot.rate_limiter` (if set)
319: by calling :meth:`telegram.ext.BaseRateLimiter.initialize`.
322: if self.rate_limiter:
323: await self.rate_limiter.initialize()
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/telegram/_bot.py | hits=1873 =====
118: from telegram.constants import InlineQueryLimit, ReactionEmoji
206: incorporated into PTB. The limitations to this argument are the same as the ones
235: ``{read, write, connect, pool}_timeout``, ``api_kwargs``. Use a named argument for those,
402: self._freeze()
691: read_timeout: ODVInput[float] = DEFAULT_NONE,
692: write_timeout: ODVInput[float] = DEFAULT_NONE,
693: connect_timeout: ODVInput[float] = DEFAULT_NONE,
694: pool_timeout: ODVInput[float] = DEFAULT_NONE,
715: read_timeout=read_timeout,
716: write_timeout=write_timeout,
717: connect_timeout=connect_timeout,
718: pool_timeout=pool_timeout,
726: read_timeout: ODVInput[float] = DEFAULT_NONE,
727: write_timeout: ODVInput[float] = DEFAULT_NONE,
728: connect_timeout: ODVInput[float] = DEFAULT_NONE,
729: pool_timeout: ODVInput[float] = DEFAULT_NONE,
744: read_timeout=read_timeout,
745: write_timeout=write_timeout,
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/aiplatform.v1beta1.json | hits=923 =====
474: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
489: "timeout": {
634: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
649: "timeout": {
671: "description": "Creates a BatchPredictionJob. A BatchPredictionJob once created will right away be attempted to start.",
890: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
905: "timeout": {
1050: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1065: "timeout": {
1376: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1391: "timeout": {
1538: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1553: "timeout": {
1696: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1711: "timeout": {
2053: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
2068: "timeout": {
2211: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/vision.v1p2beta1.json | hits=886 =====
8: "https://www.googleapis.com/auth/cloud-vision": {
15: "baseUrl": "https://vision.googleapis.com/",
17: "canonicalName": "Vision",
18: "description": "Integrates Google Vision features, including image labeling, face, logo, and landmark detection, optical character recognition (OCR), and detection of explicit content, into applications.",
20: "documentationLink": "https://cloud.google.com/vision/",
26: "id": "vision:v1p2beta1",
28: "mtlsRootUrl": "https://vision.mtls.googleapis.com/",
29: "name": "vision",
117: "id": "vision.files.annotate",
122: "$ref": "GoogleCloudVisionV1p2beta1BatchAnnotateFilesRequest"
125: "$ref": "GoogleCloudVisionV1p2beta1BatchAnnotateFilesResponse"
129: "https://www.googleapis.com/auth/cloud-vision"
136: "id": "vision.files.asyncBatchAnnotate",
141: "$ref": "GoogleCloudVisionV1p2beta1AsyncBatchAnnotateFilesRequest"
148: "https://www.googleapis.com/auth/cloud-vision"
159: "id": "vision.images.annotate",
164: "$ref": "GoogleCloudVisionV1p2beta1BatchAnnotateImagesRequest"
167: "$ref": "GoogleCloudVisionV1p2beta1BatchAnnotateImagesResponse"
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/vision.v1p1beta1.json | hits=886 =====
8: "https://www.googleapis.com/auth/cloud-vision": {
15: "baseUrl": "https://vision.googleapis.com/",
17: "canonicalName": "Vision",
18: "description": "Integrates Google Vision features, including image labeling, face, logo, and landmark detection, optical character recognition (OCR), and detection of explicit content, into applications.",
20: "documentationLink": "https://cloud.google.com/vision/",
26: "id": "vision:v1p1beta1",
28: "mtlsRootUrl": "https://vision.mtls.googleapis.com/",
29: "name": "vision",
117: "id": "vision.files.annotate",
122: "$ref": "GoogleCloudVisionV1p1beta1BatchAnnotateFilesRequest"
125: "$ref": "GoogleCloudVisionV1p1beta1BatchAnnotateFilesResponse"
129: "https://www.googleapis.com/auth/cloud-vision"
136: "id": "vision.files.asyncBatchAnnotate",
141: "$ref": "GoogleCloudVisionV1p1beta1AsyncBatchAnnotateFilesRequest"
148: "https://www.googleapis.com/auth/cloud-vision"
159: "id": "vision.images.annotate",
164: "$ref": "GoogleCloudVisionV1p1beta1BatchAnnotateImagesRequest"
167: "$ref": "GoogleCloudVisionV1p1beta1BatchAnnotateImagesResponse"
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/vision.v1.json | hits=862 =====
8: "https://www.googleapis.com/auth/cloud-vision": {
15: "baseUrl": "https://vision.googleapis.com/",
17: "canonicalName": "Vision",
18: "description": "Integrates Google Vision features, including image labeling, face, logo, and landmark detection, optical character recognition (OCR), and detection of explicit content, into applications.",
20: "documentationLink": "https://cloud.google.com/vision/",
26: "id": "vision:v1",
28: "mtlsRootUrl": "https://vision.mtls.googleapis.com/",
29: "name": "vision",
117: "id": "vision.files.annotate",
129: "https://www.googleapis.com/auth/cloud-vision"
136: "id": "vision.files.asyncBatchAnnotate",
148: "https://www.googleapis.com/auth/cloud-vision"
159: "id": "vision.images.annotate",
171: "https://www.googleapis.com/auth/cloud-vision"
178: "id": "vision.images.asyncBatchAnnotate",
190: "https://www.googleapis.com/auth/cloud-vision"
203: "id": "vision.locations.operations.get",
222: "https://www.googleapis.com/auth/cloud-vision"
===== /home/hermes/.hermes/hermes-agent/agent/auxiliary_client.py | hits=857 =====
1: """Shared auxiliary client router for side tasks (compression, search, vision, ...).
4: native Anthropic → direct API-key providers → None. Vision auto chain: main
5: provider (if a supported vision backend) → OpenRouter → Nous → Anthropic → custom.
123: # retry, so only the first per process surfaces. Separate sets let tests clear each branch.
154: # degrade to the SDK default httpx client rather than kill the job; warn once.
179: # Hermes owns aux retry/fallback policy; the SDK default (max_retries=2) would triple
181: # Hermes owns auxiliary retry + provider/model fallback policy (the same-provider transient retry in
182: # call_llm plus the except-chain fallback). The OpenAI SDK's own default (max_retries=2 → up to 3
184: # so a 120s timeout can stall ~360s before Hermes sees a single failure (issue #54465). Disable
193: # protected. Explicit host cancel (Ctrl+C, /stop) still overrides it, timeouts still fire.
199: # honors it. TIMEOUTS still fire (a hung call must die), and all OTHER aux tasks (vision, web_extract,
206: ``except Exception`` retry/fallback code never treats a host stop as a transport failure; ``cause``
220: check = _capture_aux_cancel_check()
221: return _captured_aux_cancel_requested(check) if check is not None else False
244: def _capture_aux_cancel_check() -> Optional[Callable[[], Any]]:
245: """Capture the current explicit-cancel source on the owning request thread."""
249: # Return the callable itself so attempt-local decision objects keep begin_timeout_cleanup().
254: def _captured_aux_cancel_requested(cancel_check: Callable[[], Any]) -> bool:
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/telegram/_chat.py | hits=796 =====
24: from html import escape
44: from telegram.helpers import escape_markdown
133: self._freeze()
249: return f"[{escape_markdown(name, version=2)}]({self.link})"
251: return f"[{escape_markdown(self.title, version=2)}]({self.link})"
281: return f'<a href="{self.link}">{escape(name)}</a>'
283: return f'<a href="{self.link}">{escape(self.title)}</a>'
290: read_timeout: ODVInput[float] = DEFAULT_NONE,
291: write_timeout: ODVInput[float] = DEFAULT_NONE,
292: connect_timeout: ODVInput[float] = DEFAULT_NONE,
293: pool_timeout: ODVInput[float] = DEFAULT_NONE,
308: read_timeout=read_timeout,
309: write_timeout=write_timeout,
310: connect_timeout=connect_timeout,
311: pool_timeout=pool_timeout,
319: read_timeout: ODVInput[float] = DEFAULT_NONE,
320: write_timeout: ODVInput[float] = DEFAULT_NONE,
321: connect_timeout: ODVInput[float] = DEFAULT_NONE,
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/googleapiclient/discovery_cache/documents/aiplatform.v1.json | hits=714 =====
351: "description": "Creates a BatchPredictionJob. A BatchPredictionJob once created will right away be attempted to start.",
570: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
585: "timeout": {
730: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
745: "timeout": {
1056: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1071: "timeout": {
1218: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1233: "timeout": {
1376: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1391: "timeout": {
1733: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1748: "timeout": {
1891: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
1906: "timeout": {
2053: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
2068: "timeout": {
2449: "description": "Waits until the specified long-running operation is done or reaches at most a specified timeout, returning the latest state. If the operation is already done, the latest state is immediately returned. If
===== /home/hermes/.hermes/hermes-agent/optional-skills/mlops/training/unsloth/references/llms-full.md | hits=710 =====
11: <table data-view="cards"><thead><tr><th></th><th></th><th data-hidden data-card-cover data-type="image">Cover image</th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>De
36: <table data-view="cards"><thead><tr><th></th><th></th><th data-hidden data-card-cover data-type="image">Cover image</th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td><strong>Un
42: * Unsloth is the only training framework to support all model types: [vision](https://docs.unsloth.ai/basics/vision-fine-tuning), [text-to-speech (TTS)](https://docs.unsloth.ai/basics/text-to-speech-tts-fine-tuning), BE
93: You can think of a fine-tuned model as a specialized agent designed to do specific tasks more effectively and efficiently. **Fine-tuning can replicate all of RAG's capabilities**, but not vice versa. 
103: <figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FLrqITvuoKyiMl8mqfu5B%2Flarge%20sloth%20wave.png?alt=media&token=3077792b-90ff
112: <figure><img src="https://3215535692-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxhOjnexMCB3dmuQFQ2Zq%2Fuploads%2FjT759hR4zq8ygzg1oEwI%2FLarge%20sloth%20Question%20mark.png?alt=media&token=ca
123: Minimum CUDA Capability 7.0 (V100, T4, Titan V, RTX 20 & 50, A100, H100, L40 etc) [Check your GPU!](https://developer.nvidia.com/cuda-gpus) GTX 1070, 1080 works, but is slow.
164: If you're stuck on if fine-tuning is right for you, see here! Learn about fine-tuning misconceptions, how it compared to RAG and more:
202: Fine-tuned models don’t need to call out to an external knowledge base during generation. Skipping the retrieval step means they can produce answers much more quickly. This speed makes fine-tuned models ideal for time-se
210: Even in a hybrid setup that uses both fine-tuning and RAG, the fine-tuned model provides a reliable fallback. If the retrieval component fails to find the right information or returns incorrect data, the model’s built-in
212: ## Common Misconceptions
214: Despite fine-tuning’s advantages, a few myths persist. Let’s address two of the most common misconceptions about fine-tuning:
251: There’s no single “best” approach to fine-tuning - only best practices for different scenarios. It’s important to experiment with different methods and configurations to find what works best for your dataset and use case
264: <a href="#grpo-reasoning-rl-notebooks" class="button secondary">GRPO (RL)</a><a href="#text-to-speech-tts-notebooks" class="button secondary">Text-to-speech</a><a href="#vision-multimodal-notebooks" class="button seconda
272: * [Qwen3 (14B)](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen3_\(14B\)-Reasoning-Conversational.ipynb) • [**Qwen3-VL (8B)**](https://colab.research.google.com/github/unslothai/notebooks/b
274: * [Gemma 3n (E4B)](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N_\(4B\)-Conversational.ipynb) • [Text](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3N
276: * [Gemma 3 (4B)](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_\(4B\).ipynb) • [Text](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_\(4B\).ipynb) • [V
284: * [**Qwen3-VL (8B)**](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Qwen3_VL_\(8B\)-Vision-GRPO.ipynb) - Vision **GSPO** - new
===== /home/hermes/.hermes/hermes-agent/plugins/platforms/telegram/adapter.py | hits=685 =====
41: outside multiplex — identical single-profile behavior.
57: logger.debug("Abandoned Telegram init task failed after timeout", exc_info=True)
60: async def _await_with_thread_deadline(awaitable, timeout: float, *, on_abandon=None):
61: """Wall-clock deadline that survives a blocked loop / cancellation-shielded PTB+httpcore init.
64: ``asyncio.TimeoutError`` on expiry (feeds the PTB retry ladder).
68: keeps every property the 9 call sites here rely on: thread-timer deadline that survives a blocked event
71: retry attempt, and off-loop stack-dump diagnostics when the loop never processes the expiry.
73: result = await run_bounded_async(awaitable, timeout, label="telegram-init", on_abandon=on_abandon)
75: raise asyncio.TimeoutError()
152: SEED_FALLBACK_IPS, TelegramFallbackTransport, discover_fallback_ips, parse_fallback_ip_env, tcp_keepalive_socket_options)
156: # Max seconds a send/edit may sleep inline on a flood-control RetryAfter; longer penalties fail
157: # closed with ``flood_control:{wait}`` so the caller's retry machinery owns the wait.
158: # Longer server penalties fail closed with a ``flood_control:{wait}`` SendResult so the caller's retry
159: # machinery (delivery ledger, streaming fallback) owns the wait instead of the coroutine pinning its worker
161: _FLOOD_INLINE_WAIT_CAP_SECS = 5.0
164: def _flood_cap_result(wait: float) -> "SendResult":
165: """The shared fail-closed SendResult for an over-cap flood wait."""
166: return SendResult(success=False, error=f"flood_control:{wait}", retry_after=float(wait))
===== /home/hermes/.hermes/hermes-agent/tests/test_tui_gateway_server.py | hits=665 =====
45: background ``threading.Timer`` (``_schedule_agent_build``) that calls
48: ``_make_agent`` mock, racily corrupting its captured state (the
87: thread.join(timeout=remaining)
115: # background reconnect-resume calls session.create, and an unprompted
118: # surfaces that share this cap.
131: blocked = server._ensure_active_session_slot(other, server._sessions[other])
132: assert "active session limit (1/1)" in blocked
210: result = server._methods["handoff.fail"]("r2", {"session_id": sid, "error": "late timeout"})
345: def test_prompt_submit_unknown_session_logs_warning(caplog):
357: with caplog.at_level(logging.WARNING, logger="tui_gateway.server"):
373: for rec in caplog.records
380: "method=prompt.submit" in rec.message for rec in caplog.records
407: server._sessions["iso-fallback"] = session
415: # The deferred inline-fallback thread now waits via the patient variant.
427: "id": "fallback-turn",
429: "params": {"session_id": "iso-fallback", "text": "hello"},
433: server._sessions.pop("iso-fallback", None)
437: "id": "fallback-turn",
===== /home/hermes/.hermes/hermes-agent/venv/lib/python3.11/site-packages/telegram/_message.py | hits=654 =====
25: from html import escape
29: from telegram._chatbackground import ChatBackground
101: from telegram.helpers import escape_markdown
194: self._freeze()
263: self._freeze()
296: ``{read, write, connect, pool}_timeout``, ``api_kwargs``, ``contact``, ``quote``,
343: 0-:tg-const:`telegram.constants.MessageLimit.MAX_TEXT_LENGTH` characters.
370: caption_entities (Sequence[:class:`telegram.MessageEntity`], optional): For messages with a
371: Caption. Special entities like usernames, URLs, bot commands, etc. that appear in the
372: caption. See :attr:`Message.parse_caption_entity` and :attr:`parse_caption_entities`
374: caption entities.
379: show_caption_above_media (:obj:`bool`, optional): |show_cap_above_med|
416: caption (:obj:`str`, optional): Caption for the animation, audio, document, paid media,
418: or voice, 0-:tg-const:`telegram.constants.MessageLimit.CAPTION_LENGTH` characters.
650: chat_background_set (:class:`telegram.ChatBackground`, optional): Service message: chat
651: background set.
780: 0-:tg-const:`telegram.constants.MessageLimit.MAX_TEXT_LENGTH` characters.
807: caption_entities (tuple[:class:`telegram.MessageEntity`]): Optional. For messages with a
===== /home/hermes/.hermes/hermes-agent/optional-skills/mlops/training/unsloth/references/llms-txt.md | hits=572 =====
77: We also uploaded Qwen3-Coder with native <mark style="background-color:purple;">**1M context length**</mark> extended by YaRN and full-precision 8bit and 16bit versions. [Unsloth](https://github.com/unslothai/unsloth) al
105: * <mark style="background-color:green;">**Temperature of 0.7**</mark>
109: * <mark style="background-color:green;">**Repetition Penalty of 1.05**</mark>
121: <mark style="background-color:yellow;">**Chat template for tool calling**</mark> (Getting the current temperature for San Francisco). More details here for how to format tool calls.
124: Reminder that this model supports only non-thinking mode and does not generate `<think></think>` blocks in its output. Meanwhile, specifying `enable_thinking=False` is no longer required.
133: Given that this is a non thinking model, there is no need to set `thinking=False` and the model does not generate `<think> </think>` blocks.
259: * (Optional) If multi-speaker, you could include a speaker ID token in the text or use a separate speaker embedding approach, but that’s beyond this basic guide (Elise is single-speaker).
273: If memory is very limited or if dataset is large, you can stream or load in chunks. Here, 3h of audio easily fits in RAM. If using your own dataset CSV, load it similarly.
385: | | 11 B Vision | — | [link](https://huggingface.co/unsloth/Llama-3.2-11B-Vision-Instruct-unsloth-bnb-4bit) |
386: | | 90 B Vision | — | [link](https://huggingface.co/unsloth/Llama-3.2-90B-Vision-Instruct-bnb-4bit) |
407: | **MedGemma** | 4 B (vision) | [link](https://huggingface.co/unsloth/medgemma-4b-it-GGUF) | [link](https://huggingface.co/unsloth/medgemma-4b-it-unsloth-bnb-4bit) |
408: | | 27 B (vision) | [link](https://huggingface.co/unsloth/medgemma-27b-it-GGUF) | [link](https://huggingface.co/unsloth/medgemma-27b-text-it-unsloth-bnb-4bit) |
538: | | 11 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-11B-Vision-Instruct) |
539: | | 90 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-90B-Vision-Instruct) |
657: | | 11 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-11B-Vision) | — |
658: | | 90 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-90B-Vision) | — |
692: | | 11 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-11B-Vision) | — |
693: | | 90 B Vision | [link](https://huggingface.co/unsloth/Llama-3.2-90B-Vision) | — |
===== /home/hermes/.hermes/hermes-agent/tests/agent/test_auxiliary_client.py | hits=524 =====
16: get_available_vision_backends,
17: resolve_vision_provider_client,
26: _is_rate_limit_error,
31: _try_payment_fallback,
75: # cache for later ones, causing _resolve_auto_route to skip providers
143: "model": "meta/llama-3.2-11b-vision-instruct",
149: task="vision",
151: model="meta/llama-3.2-11b-vision-instruct",
157: assert model == "meta/llama-3.2-11b-vision-instruct"
340: def test_main_agent_fallback_uses_aggregator_for_moa_main(self, tmp_path, monkeypatch):
341: """_try_main_agent_model_fallback with a moa main resolves the
343: from agent.auxiliary_client import _try_main_agent_model_fallback
353: client, model, label = _try_main_agent_model_fallback("anthropic", task="compression")
363: """_build_call_kwargs should not cap output by default (#34530).
366: model max", which is what we want for auxiliary tasks. An explicit cap only
368: max_tokens; ZAI vision rejects it entirely). The Anthropic Messages wire is
408: The ``reference_max_tokens`` config option (PR #56756) caps advisor output
410: dropped the value for OpenAI-compatible providers (PR #34845), so the cap
root@213-238-170-219:~#